Sei sulla pagina 1di 133




Section I 1

Section II 32

Section III 47

Section IV 76

Section V 87

Summary and Concluding Remarks 104

Official Text of Questions and Consequent Changes 112

It is a pity that the IIT’s have still not adopted the practice of making the
JEE question papers public immediately after the examination. So, these
comments were based on memorised versions of the questions. After about
two months, an official text of the mathematics question paper was obtained
using the Right to Information Act. As a result, some of the answers and
comments had to be revised. All these changes have been listed at the end.
The opinions expressed are the author’s personal ones and all references are to
the book Educative JEE Mathematics by the author, unless stated otherwise.
The question paper is divided into five sections. Section I (Q. 1 to 12) con-
sists of multiple choice questions with only one correct answer. Each question
carries (3, −1) points, meaning that 3 points are given for a correct answer
while −1 point (i.e. one negative point) is given for each wrong answer.
Section II (Q. 13 to 20) consists of multiple choice questions where one or
more answers are correct, each question carrying (5, −1) points. Section III
(Q. 21 to 32) has four passages, each passage carrying three multiple choice
questions with only one correct answer, each question having (5, −2) points.
Section IV (Q. 33 to 36) has four questions, where numerical answers have
to be filled in and each question has (6, 0) points. The last section, Section
V (Q. 37 to 40) has four questions, each asking for matching the pairs and
carrying (6, 0) points.
The author will be happy to receive comments by e-mail (
Corrections/additions will be made from time to time and displayed here.

Only one of the given answers is correct.

Q. 1 If t1 = (tan θ)tan θ , t2 = (tan θ)cot θ , t3 = (cot θ)tan θ and t4 = (cot θ)cot θ ,

where θ ∈ (0, π/4), then
(A) t4 < t2 < t1 < t3 (B) t4 < t1 < t3 < t2
(C) t4 < t3 < t2 < t1 (D) t2 < t1 < t3 < t4

Answer and Comments: (D). Superficially, this is a problem in

trigonometry. But in reality it is a problem about inequalities, specif-
ically, about relating inequalities of powers to those of the bases and
of the exponents. Let α = tan θ. Then cot θ = α1 . As θ is given to
lie between 0 and π4 we have 0 < α < 1. Conversely, every positive
real number α ∈ (0, 1) can be expressed as tan θ for some (unique)
θ ∈ (0, π4 ). So translated in terms of α, the problem amounts to asking
you to arrange the four real numbers αα , α1/α , ( α1 )α and ( α1 )1/α in the
ascending order, given that 0 < α < 1. The first step is to recognise
that since α is positive and less than 1, so are all powers of α in which
the exponents are positive. Similarly, α1 > 1 and therefore all its powers
with positive exponents are also greater than 1. Thus we immediately
have that t1 , t2 are both less than 1 while t3 , t4 are both greater than
1. (Both these statements are obvious when the exponents are positive
integers. For the general case, it is better to take logarithms. Thus,
if a > 1, then ln a > 0 which makes b ln a > 0 whenever b > 0. But
b ln a = ln(ab ) and hence ab > 1. On the other hand, for 0 < a < 1, we
have ln a < 0 and therefore ln(ab ) = b ln a < 0 for b > 0, which gives
ab < 1.)
It now only remains to compare t1 with t2 and t3 with t4 . The
latter comparison is easier. When the bases are the same and greater
than 1, the powers are in the same order as the exponents. In symbols,
if a > 1 and b1 < b2 then ab1 < ab2 . Once again, this appears obvious.
But a rigorous proof can be given using logarithms, because ln(ab1 ) =
b1 ln a < b2 ln a = ln(ab2 ).) Now, since α1 > 1, we also have α < 1/α.

Therefore, by what we just said, ( α1 )α < ( α1 )1/α . Or, in the notation of
the given problem, (cot θ)tan θ < (cot θ)cot θ , i.e. t3 < t4 .
But the comparison between t1 and t2 , i.e. between αα and α1/α is
a little tricky. Here the base α is less than one. As a result, the powers
are in the opposite orders as the exponents. In symbols, if 0 < a < 1
and b1 < b2 then ab1 > ab2 . Once again the best way to see this is to
take logarithms, keeping in mind that ln a < 0. (Note that when we
talk of powers, we tacitly assume that the bases are positive. Also, the
proofs above are independent of whether we take logarithms w.r.t. the
base e or w.r.t. some other base as long as that base is greater than
1. If the base of the logarithms is taken to be positive but less than 1,
then x < y is equivalent to log x > log y and not to log x < log y. As
a result, the inequalities in the proofs above will be reversed. But the
ultimate results will not change. That is, powers with equal bases are
related the same way as their exponents are when the base is greater
than 1 and oppositely when it is less than 1.)
So, from 0 < α < 1 and α < 1/α, we get αα > α1/α . In terms of
the given notations, t1 > t2 . Putting this together with t3 < t4 and the
fact t1 < t3 (since t1 < 1 and t3 > 1), we get t2 < t1 < t3 < t4 , i.e. (D)
as the right answer.
The essential facts needed from trigonometry are only that tan θ
and cot θ are reciprocals of each other and that the former lies between
0 and 1 for the given values of θ. The crux of the problem lies in the
comparisons of powers with common bases. When the common base
is greater than 1, they behave in the expected way. But one has to be
wary when the common base is positive but less than 1. A person who is
careful enough to realise can be safely credited to know the reasons for
doing so, without having to spell them out. That makes this problem
ideal to be asked as a multiple choice question. It is reported that
the problem appears in the book 103 Trigonometry Problems by Titu
Andreescu and Zuming Feng, a Birkhauser Publication.
Q. 2 For x > 0, lim (sin x) x + ( )sin x equals
x→0 x
(A) 0 (B) −1
(C) 1 (D) 2

Answer and Comments: (C). A most natural approach is to write
the given function, say f (x) as the sum of two functions, say f1 (x) =
(sin x)1/x and f2 (x) = ( x1 )sin x and consider the separate limits of each
of these two functions as x tends to 0 from the right. If both these
limits exist then their sum will give us the desired limit.
So, let us first consider lim+ (sin x)1/x . Here the expression is a
power and as x approaches 0 (from the right), the base sin x tends to 0
while the exponent 1/x tends to ∞. So it is clear that the expression
(sin x)1/x tends to 0 as x → 0+ . (For a formal proof, we can take its
ln sin x
log and consider the limit of the log. We have ln((sin x)1/x ) = .
ln sin x sin x
Rewrite this as × . The second factor tends to 1 as x tends
sin x x
to 0. So we need to consider only the limit of the first factor. Writing y
ln y
for sin x, this is the same as lim+ . As the numerator tends to −∞
y→0 y
while the denominator tends to 0+ , it is clear that the ratio tends to
ln sin x
−∞. So lim+ ln((sin x)1/x ) = = −∞ and therefore the original
x→0 x
limit, viz. lim+ (sin x)1/x equals 0.)
Let us now tackle the second limit, viz. lim+ ( )sin x . This is an
x→0 x
indeterminate form of the ∞0 type. As the expression is a power, we
once again convert the problem by taking logs. We have ln(( x1 )sin x ) =
sin x ln( x1 ) = − sin x ln x. Once again we divide and multiply by x to
sin x
rewrite this as − (x ln x), which enables us to focus our attention
on lim+ x ln x. Here the first factor tends to 0 while the second tends to
−∞. So, this is again an indeterminate form of the 0×∞ type. To find
ln y
the limit we put y = 1/x and write the limit as y→∞
lim − . Here the
numerator and the denominator both tend to ∞. But it is well-known
that the logarithm of y tends to ∞ much slower than y (in fact, slower
than any power y α , as long as the exponent α is positive). This is a
consequence of the fact that any polynomial growth is slower than an
exponential growth, see Exercise (6.51). Or, one can apply L’Hôpital’s
1/y 1
rule to convert the limit to y→∞
lim = y→∞
lim .
1 y

1 1
So, we have proved that lim+ ln(( )sin x = 0 and therefore lim+ ( )sin x
x→0 x x→0 x
equals e0 i.e. 1. This gives us the limit of the second term in the given
problem. As the limit of the first term is 0, the final answer is 1.
Although we have given the reasoning rather elaborately, most
of the thinking is intuitive. That the limit of the first term is 0 is
clear from the fact that the base gets smaller and tends to 0 while the
exponent gets larger and tends to ∞ as x tends to 0 from the right.
The calculation of the limit of the second term is a little subtle. But
the technique of taking logarithms is a very common one while dealing
with limits of powers. In a compact form, it says that the limit of a
log is the log of the limit. The theoretical justification is based on the
continuity of the exponential function, as pointed out at the beginning
of Comment No. 8 of Chapter 15. The trick of replacing sin x by x is
also used frequently. (See Comment No. 6 of Chapter 15.) In fact, after
gaining some practice, these things can be done mentally. The really
non-trivial fact needed in the evaluation of the second limit is that
x ln x → 0 as x → 0+ . This is quite well known and is often expressed
by saying that logarithmic infinity is weaker than an algebraic infinity.
In essence, the problem involves the evaluation of two separate
limits and adding them. Not all problems about the limit of a sum are
so straightforward. It may happen that neither lim f1 (x) nor lim f2 (x)
x→c x→c
exists and still lim(f1 (x) + f2 (x)) exists. For example, take f1 (x) = sin1 x
and f2 (x) = − x1 and c = 0. In such cases, the simple-minded method
we applied for the present problem does not work and more delicate
methods have to be used. (For the particular example just given, see
Comment No. 7 of Chapter 15.)
Finally, it may happen that one of the two limits exists and the other
does not. In that case then the limit of the sum will not exist. (The
proof is basically the same as that of Exercise (15.20).) The problem
could have been a little more revealing by giving the non-existence of
the limit as one possible option. Such an option is likely to tempt those
who are not very scrupulous.

Q. 3 One
√ angle of an isosceles triangle is 120 and the radius of its incircle
is 3 units. Then the area of the triangle in sq. units is

√ √
(A) 7 + 12√ 3 (B) 12 − 7 3
(C) 12 + 7 3 (D) 4π

Answer and Comments: (C). There are several formulas which ex-
press the area ∆ of a triangle ABC in terms of its inradius r and
something else. The most well-known of these is

∆ = rs (1)

where s is the semi-perimeter of the triangle. In the present problem,

this is not of much direct use, because although we are given r, we are
not given the sides. So, to find the semi-perimeter, we shall have to
first find the sides from the angles and the inradius. This can indeed
be done and yields the following formula which gives the area ∆ of a
triangle ABC most directly in terms of the inradius r and the angles.
∆ = r 2 (cot + cot + cot ) (2)
2 2 2
(see Equation (1) in the solution to the Main Problem of Chapter 11).
In the present
√ problem, if we take 6 A = 120◦ , then 6 B = 6 C = 30◦ .
As r = 3, we have ∆ = 3(cot 60◦ + 2 cot 15◦ ). The value of cot 60◦
is standard and equals tan 30◦ = √13 . The value of cot 15◦ is not so
standard but can be calculated easily from tan 30◦ . Call tan 15◦ as α.
Then we have
1 2α
tan 30◦ = √ = (3)
3 1 − α2

This can be √cast as a quadratic in α, viz.√α2 + 2 3α − 1 = 0, which
gives α = − 3 ± 2 and hence α = 2 − 3 since tan 15◦ > 0. Thus
1 √
cot 15◦ = √ = 2 + 3 after rationalisation. Putting these values
√2 − 3 √ √
and r = 3 in (2) gives ∆ = 3( √13 + 4 + 2 3) = 12 + 7 3 sq. units.
Although we have calculated the value of cot 15◦ by solving a
quadratic, some standard textbooks list the values of sin 15◦ and cos 15◦
among the standard ones. A student who remembers them correctly
will save some time. In that case the method given here is the quickest
solution to the problem.

What if you cannot recall formula (2), which is, after all, not so
standard as (1)? Even then, everything is not lost. A solution based
directly on (1) (i.e. bypassing (2)) is also possible. Note that since the
angles of the triangle ABC are known, we already know the relative
proportions of its sides. Therefore each side can be expressed as a
multiple of any one of the sides, say a. In that case, the semi-perimeter
s can also be expressed in terms of a. We are already given the value
of r. As a result, using (1) we can now express ∆ in terms of a.
To work out the details,
from the figure we see that A
◦ ◦
a = 2b cos 30 = 2c cos 30 , c 60 60 b
30 30
a a /2 a /2
b=c= √ (4) B D C
(This can also be obtained by applying the sine rule to ∆ABC.)
From (4) we get,
1 1 1
s = (a + b + c) = ( + √ )a (5)
2 2 3
Therefore from (1) we have

1 1 3+2
∆= 3 +√ a= a (6)
2 3 2

(An alert reader will notice something amiss here. The L.H.S. is an
area and hence the square of a length while the R.H.S. is a length.
This happens
√ because we have replaced r, which is a length, by the
scalar 3.)
We are still not at the answer. To get to it from (6) we need to
know the value of a. We are given the value of r. If we can write r in
terms of a, then we shall get the value of a and hence that of ∆.
There are indeed formulas for expressing the inradius r in terms
of the sides of a triangle. But the most standard formula of this type

is (1) itself recast slightly, viz. r = . Obviously, this will take us to a
vicious cycle where we have to know ∆ to get a and to get ∆ we have

to know a. Here again, a less well-known formula can do the trick. One
such formula is
r = (s − a) tan = (s − b) tan = (s − c) tan (7)
2 2 2
Obviously, in the present problem, it is easier to apply this with A
instead of B or C. From (5) we get

1 1 ◦ 2− 3
r = ( √ − )a tan 60 = ( )a (8)
3 2 2
√ √
2 √3
But we are already given that r = 3. So, from (8) we get a = 2− 3
√ √
Hence from (5) we get s = ( 12 + √13 ) 2−
2 √3
= (2− 3)
. As we already know
√ √ √
r = 3, (1) now gives ∆ = 2+ √3 3 which gives the same answer as
2− 3
before upon rationalisation.
There is one more way out which does not use the relatively obscure
formula (7). Equation (6) is an equation in the two unknowns ∆ and a.
If we can get some other equation relating these two unknowns, then
solving it simultaneously with (6) we can get the value of ∆ (in which
we are interested) and also that of a (in which we are not interested
per se).
So, we look for another equation involving ∆ and a. This can be
done using either of the two other standard formulas for ∆, viz.
1 1 1
∆ = ab sin C = bc sin A = ca sin B (9)
q 2 2
or, ∆ = s(s − a)(s − b)(s − c) (10)
We can now put (4) and (5) into (10). Or we can put (4) into (9) and
use the fact that we know the angles A, B, C. Opting for the second
method, we get
a2 a2
∆= sin 120◦ = √ (11)
6 4 3
As we are not interested in a, we eliminate a between (6) and (11) to
2∆ √
√ = 4 3∆ (12)

As ∆ 6= 0, we cancel it from both the sides and get
√ √ √ √ √
∆ = 3( 3 + 2)2 = 3(7 + 4 3) = 12 + 7 3 (13)
which is the same answer as before.
Although the problem is a simple problem about solutions of
triangles, it amply illustrates the good as well as the bad effects of the
rich variety of formulas about triangles. For example, we listed four
formulas, viz. (1), (2), (9) and (10), for the area of a triangle. If you
use the right formula (which is (2) for the present problem) you get the
answer almost instantaneously. Otherwise you may wind up working
in circles. But if you analyse the failure, then the situation can be
Q. 4 Suppose a, b, c are the sides of a triangle and no two of them are equal.
Let λ ∈ IR. If the roots of the equation x2 + 2(a + b + c)x + 3λ(ab +
bc + ca) = 0 are real, then
(A) λ < 34 (B) λ > 53
(C) λ ∈ ( 31 , 53 ) (D) λ ∈ ( 34 , 35 )

Answer and Comments: (A). As with many other problems in JEE,

here a problem from one area is artificially combined with some tidbit
from another. The garb of quadratic equations here is quite superficial
and easy to lift if we see that the given condition is merely a twisted way
of giving the inequality (a + b + c)2 ≥ 3λ(ab + bc + ca) or equivalently,
(a + b + c)2
λ≤ (1)
3(ab + bc + ca)
(The equivalence follows because a, b, c and hence ab + bc + ca are all
Thus the problem is reduced to a problem of trigonometric optimi-
sation, viz. finding the range of the possible values of the expression
(a + b + c)2
where a, b, c are the sides of a scalene triangle (i.e. a
3(ab + bc + ca)
triangle in which no two sides are equal). In essence, this is the same
as the 1979 JEE problem, given in Exercise (14.4). It says that for any
triangle ABC, we must have
(a + b + c)2
3≤ ≤4 (2)
ab + bc + ca

Further, equality holds on the left only when the triangle is equi-
lateral. Although not given in the exercise, equality holds on the
right only for degenerate triangles where one of the angles is 0. (In
such a case, in the middle expression, the other two sides have to be
equal by the triangle inequality. Or, we can replace it by the ratio
(sin A + sin B + sin C)2
which makes sense and equals 4
sin A sin B + sin B sin C + sin C sin A
even when one of A, B, C is 0 while the other two are positive and add
up to π.)
As we assume that the given triangle is not degenerate, it follows
immediately from (1) and (2) that the value of the real number λ is
less than 34 . Hence (A) is a correct alternative. It may be argued that
when λ < 34 , we also have λ < 35 and therefore (C) is also a correct
answer. But this claim is fallacious because even if λ ≤ 13 , the quadratic
equation given in the statement of the problem will have real roots. So
the statement of the problem does not logically force λ to lie in the
open interval ( 31 , 43 ).
There is, however, some impropriety in the problem. The coefficient
3 of λ in the statement of the problem has absolutely no role, except
to unnecessarily complicate the problem a little. More seriously, the
hypothesis that no two sides of the triangle are equal is nowhere needed
in the solution. The problem basically deals with the second, and not
the first, inequality in (2) where, to ensure strictness of the inequality all
we need is that the triangle be non-degenerate. It makes no difference
whether any of the sides are equal or not. Even if the inequality on the
left in (2) was relevant in the problem, for its strictness all you need is
that a, b, c are not all equal and not that no two of them are equal.
It appears that the problem has been designed by taking the 1979
JEE problem stated above and putting the garb of quadratic equations
on it, The impropriety just mentioned suggests that this conversion
was done without giving much thought. (See the remarks at the end of
Comment No. 4 of Chapter 12 about how hastily converted problems
sometimes result in comic situations.)

Q. 5 If 0 < θ < 2π, then the interval(s) of values for which 2 sin2 θ − 5 sin θ +
2 > 0, is

(A) (0, π6 ) ∪ ( 5π6
, 2π) (B) ( π8 , 5π
(C) (0, π8 ) ∪ ( π6 , 5π
) (D) ( 41π
, π)

Answer and Comments: (A). Another straightforward problem,

where trigonometric equations are combined with a little bit about
the sign of a quadratic expression. Call sin θ as t.√ The roots of the
5± 9
quadratic equation 2t2 − 5t + 2 = 0 are t = , i.e. t = 2 and
t = 12 . Also the leading coefficient 2 is positive. So the expression
2t2 − 5t + 2 is positive for all values of t outside the two roots (see Com-
ment No. 11 of Chapter 2), i.e. for t ∈ / [ 12 , 2]. So the given inequality
is satisfied if and only if sin θ < 12 or sin θ > 2. The second possibility
is vacuous because sin θ can only take values between −1 and 1. So
the problem now is to decide for which values of θ ∈ (0, 2π), we have
sin θ < 21 . Keeping in mind that for θ ∈ (0, 2π), sin θ = 12 for θ = π6 and
for θ = 5π6
and the increasing/decreasing behaviour of the graph of the
sine function, this set comes out to be the union of the intervals (0, π6 )
and ( 5π
, 2π).
Q. 6 If w = α+iβ where β 6= 0 and z 6= 1 satisfies the condition that 1−z
is purely real, then the set of values of z is
(A) {z : |z| = 1} (B) {z : z = z}
(C) {z : z 6= 1} (D) {z : |z| = 1, z 6= 1}

Answer and Comments: (D). A sneaky way to answer this problem

is by eliminating the incorrect alternatives. (A) is nipped in the bud
because the set {z : |z| = 1} contains the point 1 which is specifically
excluded by the statement of the problem. To rule out (B) and (C) it
suffices to show that the respective
 sets contain some points z which
do not satisfy the condition that w−wz
is purely real. By inspection,
z = 0 is one such point because for z = 0, this expression becomes w
(= α + iβ) which is not real since β is given to be non-zero. This single
value eliminates both (B) and (C) simultaneously. As one of the given
answers has to be correct, it must be (D).
Although good enough (and indeed recommended in order to save
time) for a multiple choice examination, this solution can hardly be
defended from an educative point of view. So let us see if there is some

‘honest’ way to arrive at the answer. As usual, we denote a typical
complex number z as x + iy where x, y are real. Now w = α + iβ is
w − wz
a fixed complex number. Let us denote the complex number
by Z = X + iY . The problem asks us to identify those z 6= 1 for
which the corresponding Z is real. (The word ‘purely’ is added only for
emphasis. ‘Purely real’ is the same as ‘real’. Although the expression
‘purely imaginary’ is used commonly, ‘purely real’ is not so commonly
used. A candidate who sees it for the first time in an examination may
get a little confused.)
A brute force way of doing this would be to work in terms of
the real numbers x, y, X and Y instead of the complex numbers z and
Z. Thus we express X and Y in terms of x and y and then determine
the conditions on x and y under which Y is zero. A straightforward
calculation gives

w − wz
X + iY = Z =
(α + iβ) − (α − iβ)(x + iy)
(1 − x) − iy
(α − αx − βy) + i(β − αy + βx)
(1 − x) − iy
((α − αx − βy) + i(β − αy + βx))((1 − x) + iy)
= (1)
(1 − x)2 + y 2

As we are interested only in the imaginary part, viz. Y of Z, we

need not expand the numerator of the R.H.S. fully. We take only its
imaginary part and set it equal to 0. Thus we get that Z is real if and
only if

(β − αy + βx)(1 − x) + (α − αx − βy)y = 0 (2)

which represents a circle because the coefficients of x2 and y 2 are equal

(viz. −β each) and non-zero and there is no xy-term. So if we want
to abandon the honest approach half way, we have already narrowed
the choice to (A) and (D). But since (A) contains the forbidden point
z = 1, we choose (D) as the correct answer.

But let us follow the honest approach all the way through and
identify the circle represented by (2). Upon simplification, (2) can be
rewritten as

β(1 − x2 − y 2 ) = 0 (3)

Canceling β which is given to be non-zero, this is simply x2 + y 2 = 1,

i.e. the equation of the unit circle, which in the complex form is |z| = 1.
But since we must exclude the point (1, 0), the correct answer is (D)
and not (A).
Although the computations in the solution above are not pro-
hibitive, while tackling a problem about complex numbers, it is always
a good idea to first see if the complex numbers can be handled using
the constructions peculiar to complex numbers, rather than routinely
translating everything in terms of their real and imaginary parts. Just
as problems of pure geometry can sometimes be solved more elegantly
using methods of pure geometry than by brute force conversion to co-
ordinates, the same thing holds for problems about complex numbers.
The present problem is a good illustration of this. The requirement
that Z be real was translated above to mean that Y = 0. But we can
also express this requirement in terms of complex conjugates. Thus Z
is real if and only if it coincides with its complex conjugate, i.e. if and
only if Z = Z. The advantage of this approach is that the complex con-
jugation preserves sums, differences, products and quotients. Therefore
if we have an expression for Z which involves only these basic opera-
tions, then an expression for Z can be written down instantaneously.
The present problem is of this type. We are given that
w − wz
Z= (4)
Therefore, by properties of complex conjugates, we have
w − wz
Z= (5)
The requirement that Z be real now translates as a requirement about
z, viz.,
w − wz w − wz
= (6)
1−z 1−z

Cross-multiplication gives

w + wzz = w + wzz (7)


w − w = (w − w)zz (8)

Since w = α + iβ, we have w − w = 2β. So, (8) becomes

2β = 2βzz (9)

Now for the first time we use the fact that β is non-zero. This gives
zz = 1. Thus the point z lies on the unit circle. Note, however, that
z 6= 1, because for z = 1, Z is not even defined. So the correct answer
is (D) and not (A).
Note that the equations (3) and (9) are essentially the same. But the
manner in which they are arrived at is different. (3) was obtained by a
brute force conversion of a complex number to an ordered pair of real
numbers, while (9) came out elegantly because of the handy properties
of complex conjugation. Obviously, the latter approach is far better.
Unfortunately, the wording of the problem is a little likely to mislead
a candidate. There was absolutely no need to introduce the real and
imaginary parts, viz. α and β respectively, of the complex number w.
Instead of saying that β 6= 0, it would have been fine to simply say
that w is not real. That is equivalent to saying that w − w 6= 0 and so
the desired conclusion, viz. zz = 1 would have followed directly from
(8). (The unscrupulous will anyway cancel (w − w) from both the sides
of (8) without bothering to check if it is non-zero.) By unnecessarily
introducing β (and α too), a candidate is tempted to convert the whole
problem in terms of the real and imaginary parts.
In this problem, Z was a complex valued function of the com-
w − wz
plex variable, given by Z = . This function is an example
of what is called a linear fractional transformation or a Mobius
transformation. More generally, the term is applied to any transfor-
az + b
mation, say T of the form T (z) = where a, b, c, d are complex
cz + d
numbers. We generally assume that ad − bc 6= 0 as otherwise the

transformation degenerates into a constant. (In the present problem,
a = −w, b = w, c = −1 and d = 1. The requirement ad 6= bc translates
into w 6= w.)
Such a Mobius transformation is defined at all points of the
complex plane except where its denominator vanishes, i.e. except the
point −d/c. It is easy to show that this transformation is one-to-one
and almost onto in the sense that its range consists of all points of
the Z-plane except the point a/c. Further it is easy to show that
the inverse transformation of a Mobius transformation is also a Moius
transformation. In fact, if Z = az+b
is a Mobius transformation, then
the inverse Mobius transformation is given explicitly by z = −dZ+b
In the present problem we are given a Mobius transformation
w − wz
Z = T (z) = and are asked to find the inverse image, under T ,
of the real-axis in the Z-plane. The real axis is a straight line. But its
inverse image in the z-plane came out to be a circle (except one point).
Note further that the real axis in the Z-plane divides it into two halves,
the ‘upper’ half plane where the imaginary part of Z is positive and
the ‘lower’ half plane. It is not hard to show that the inverse image of
one of these is the disc inside the unit circle in the z-plane while the
inverse image of the other is the exterior of this circle. (Which one goes
to which will depend upon the sign of β.)
This is typical of all Mobius transformations. They take straight
lines to either circles or to other straight lines and the same holds for
circles. And accordingly, they take discs to discs or to half planes and
half planes to discs or half planes. Exactly which possibility holds de-
pends on the values of the constants a, b, c, d and also on the particular
region. If c = 0, then nothing very strange happens. Straight lines go
to straight lines and circles to circles. But if c 6= 0 (as is the case in the
present problem) then such strange things do happen and they can be
used advantageously. Just as in the evaluation of definite integrals we
sometimes change the interval of integration to a more convenient one
by a suitable substitution (accompanied, of course, by a corresponding
change of the integrand too), sometimes in dealing with certain prob-
lems in the complex plane, we want to convert a disc domain to a half
plane. A suitable Mobius transformation is employed to do the trick.
The lacunae created by the vanishing of the denominator can be

patched up by adding an extra point, denoted by ∞ (and called, simply,
the point at infinity). This point is common to all straight lines. If we
include that as a part of the real axis in the Z-plane in the present
problem, then its inverse image will be the entire unit circle, including
w − wz
the point z = 1 because the transformation Z = maps 1 to
∞. Note also that with the inclusion of this point in the domain, every
az + b
Mobius transformation becomes onto, because if T (z) = , then
cz + d
we can set T (∞) = .
These things are, of course, beyond the JEE syllabus. But once in a
while the JEE papers contain problems whose origin lies in something
from the outside the syllabus. For example, the 1998 JEE problem
given in Exercise (21.27) is based on a property of quaternions given
in Exercise (21.26). Similarly, the JEE 2005 Mathematics papers in-
cluded a problem based on lattices in the plane and also a problem
based on the Cayley Hamilton equation of a matrix. (See the author’s
educative commentary on the JEE 2005 Mathematics Papers.) Al-
though such problems can always be done, and indeed are expected to
be done, without a knowledge of the more sophisticated concepts or
theorems, they can be truly appreciated only with some idea of where
they originate.
Q. 7 If r, s, t are prime numbers and p, q are positive integers such that the
L.C.M. of p, q is r 2 s4 t2 , then the number of ordered pairs (p, q) is
(A) 224 (B) 225
(C) 252 (D) 256

Answer and Comments: (B). This is a simple problem in number

theory, once you understand what it really is. If the given setting is
too abstract for you, it is a good idea to work out some special cases
as illustrations (for example, by taking r, s, t to be the smallest three
primes, viz. 2, 3, 5 respectively, in which case r 2 s4 t2 = 8100). Once you
solve the problem for this special case you will realise that the method
is quite independent of which three particular primes are represented
by r, s, t. With a little more maturity, you can see this instinctively and
proceed with the general problem as we now do. But before tackling it
honestly, we observe that there is an unwarranted short cut to eliminate

the incorrect alternatives. The desired ordered pairs (p, q) are of two
types: those in which p = q and those in which p 6= q. Clearly, there is
only one pair of the first type, because the only way the l.c.m. of two
equal numbers can be a given number is when both of them are equal
to that number. For every ordered pair of the second type, viz. (p, q)
where p 6= q, the pair (q, p) is also to be counted ans is distinct from
(q, p). So, without actually counting, the number of pairs of the second
type is even. Therefore the total number of desired pairs is odd and
since this happens only for (B), if at all one of the answers is correct,
it has to be (B).
Now, for an honest solution, we are told that p and q are two
positive integers whose l.c.m. is r 2 s4 t2 . This first of all means that
neither p nor q can have any prime factor besides r, s and t. So each
of them is a product of powers of some of these three primes. We can
therefore write p, q in the form

p = r a sb tc and q = r u sv tw (1)

where a, b, c, u, v, w are non-negative integers. Then the l.c.m., say e,

of p and q is given by

e = r i sj tk (2)


i = max{a, u}, j = max{b, v} and k = max{c, w} (3)

This is the key idea of the problem. The problem is now reduced to find-
ing the number of triplets of ordered pairs of the form {(a, u), (b, v), (c, w)}
where a, b, c, u, v, w are non-negative integers that satisfy

max{a, u} = 2, max{b, v} = 4 and max{c, w} = 2 (4)

Let us see in how many ways the first entry of this triplet, viz.,
(a, u) can be formed. We want at least one of a and u to equal 2. If we
let a = 2, then the possible values of u are 0, 1 and 2. These are three
possibilities. Similarly, with u = 2 there will be three possibilities, viz.
a = 0, 1 or 2. So, in all the first ordered pair (a, u) can be formed in 6
ways. But the possibility (2, 2) has been counted twice. So, the number

of ordered pairs of the type (a, u) that satisfy the first requirement in
(4) is 5 and not 6.
By an entirely analogous reasoning, the number of ordered pairs of
the form (b, v) which satisfy the second requirement in (4) is 2 × 5 −
1, i.e. 9 while that of ordered pairs of the type (c, w) satisfying the
third requirement in (4) is 5. But the ways these three ordered pairs
are formed are completely independent of each other. So the total
number of triplets of ordered pairs of the form {(a, u), (b, v), (c, w)}
where a, b, c, u, v, w are non-negative integers that satisfy (4) is 5 × 9 ×
5 = 225. Hence (B) is the correct answer.
The number theory involved in the problem is very elementary.
Thereafter it is essentially a counting problem. Although the reasoning
takes a long time to write down, once you hit the essential idea it does
not take much time to work out the details mentally. That makes this
problem ideal as a multiple choice question. It is, in fact, a very good
One of the common pitfalls in reasoning in this problem is to
count the ordered pairs (2, 2), (4, 4) and (2, 2) twice each. In that case
the answer would come out to be 6 × 10 × 6 = 360. As this is not
given as a possible answer, a good student is alerted that he is making
some mistake. If the paper-setters have done this intentionally, it is
commendable on their part because it shows that they are trying to
help a good student rather than pounce on his single weakness when
he has done most of the work correctly. This is important because
unlike in a conventional examination, where you can get some partial
credit, in a multiple choice question, a silly slip is even more fatal than
total inability to solve a problem. It costs you heavily both in terms
of the time spent and the negative credit you earn in spite of doing
most of the work correctly. However, if the paper-setters are really so
concerned about a sincere student, they ought to have included at least
one fake answer with an odd number to preclude the sneaky short cut
given at the start.
x2 − 1
Q. 8 √ dx equals
x3 2x4 − 2x2 + 1

√ √
2x4 − 2x2 + 1 2x4 − 2x2 + 1
(A) +c (B) +c
2x2 x3
√ √
2x4 − 2x2 + 1 2x4 − 2x2 + 1
(C) +c (D) +c
x x2

Answer and Comments: (A). This is evidently a problem about

finding an antiderivative of a given function. Had it been asked in
the conventional form, then one would really have to find it. But the
multiple choice format obviates the need to do so. If one wants, one
can simply differentiate each of the given alternatives and see which
derivative equals the given integrand. Moreover, as only one answer is
correct, the search stops as soon as you have found one match. As fate
would have it, the way the answers are ordered in the present problem,
the derivative of the very first one tallies with the integrand. So, (A)
is the right answer. (Sometimes, to reduce the chances of copying, the
alternatives are shuffled among themselves in the various versions of
the same question. In that case, if in a problem like this, the correct
alternative is listed as (A) in some question papers and as (D) in some
other, that means between two crooks one is luckier than the other
because unless the crook is extra smart, he would try the answers one
by one from (A) to (D) till he gets the right one!)
Instead of differentiating each alternative separately, a smart crook
can concentrate on their similarity. Forgetting the constant
√ of inte-
2x − 2x2 + 1
gration c, each of the four alternatives is of the form
where u(x) is some function of x. If we differentiate this, the derivative,
after a little simplification, comes out to be
(4x3 − 2x)u(x) − (2x4 − 2x2 + 1)u′(x)
√ (1)
u(x)2 2x4 − 2x2 + 1
This will match with the integrand if and only if

(x2 − 1)(u(x))2 = (4x3 − 2x)x3 u(x) − u′ (x)x3 (2x4 − 2x2 + 1) (2)

The choices given for the answer correspond to u(x) = 2x2 , x3 , x and
x2 respectively. Of these four, the possibilities u(x) = x3 and u(x) =
x are ruled out by considerations of the degrees of the two sides as

polynomials in x. The case u(x) = x2 is dismissed by comparing the
leading coefficients of both the sides.
Let us now leave aside these crooks and get the answer honestly,
x2 − 1
i.e. by finding an antiderivative of 3 √ 4 . Note that ev-
x 2x − 2x2 + 1
erywhere we have only even powers of x except in the factor x3 in the
1 x
denominator. We can write 3 as 4 and combine the x in the numer-
x x
ator nicely with dx so that it becomes 12 d(x2 ). This suggests that the
substitution u = x2 may work. Trying it, the integral, say I, becomes
1 u−1
I= √ du (3)
2 u2 2u2 − 2u + 1
It is now tempting to get rid of the radical by calling
√ the radical itself as
a new variable, say v, i.e. by substituting v = 2u − 2u + 1. If we do
(2u − 1)du
so, then the expression for dv becomes √ . So, this would have
been just the substitution we need had the numerator in the integrand
in (3) been 2u − 1 instead of u − 1 and had there been no u2 in the
denominator. But unfortunately, we are stuck with these.

Let us now see if we can transform the radical q2u2 − 2u + 1 to a
more manageable radical, say, a radical of the form f (u) where f (u)
is some function of u. The integrand in (3) would then look like q
f (u)
where g(u) is also some function of u. In fact, g(u) depends on f (u).
More specifically,
(u − 1) f (u)
g(u) = 2 √ 2 (4)
u 2u − 2u + 1
By a ‘more manageable’ radical f (u) we mean that if we put z =
q f ′ (u)du
f (u), then the expression for dz, viz. should be equal to
g(u) except possibly for some constant factor. If this happens then the
substitution z = f (u) will work. In simpler terms, we would like g(u)
to be a constant multiple of f ′ (u).

Let us now look for this magic radical f (u). As the things
already stand, in the integrand in (3), we have f (u) = 2u2 − 2u + 1 and
g(u) = 2
. Here f ′ (u) = 4u − 2 and g(u) is not a constant multiple
of f ′ (u). So this choice of f (u) is no good as we already know anyway.
But let us take out a factor u2 from this f (u) and pass it to g(u). So
2 1 u−1
our new f (u) now equals (2 − + 2 ) and our new g(u) is . This
u u u3
2 2 2(u − 1)
time f ′ (u) = 2 − 3 = which is indeed a constant multiple
u u u3
of g(u).
So, we have hit the right choice. Calculating the antiderivative is
now a clerical matter. Continuing from (3) we have
1 u−1
I = √ du
2 u 2u2 − 2u + 1

1Z u−1
= s du (5)
2 2 1
u3 2 − + 2
u u
2 1
We now put z = 2− + 2 . Then we have
u u
2 2
− 3 (u − 1)
dz = s u u du = s du (6)
2 1 3
2 1
2 2− + 2 u 2− + 2
u u u u

Putting (5) and (6) together, we get

1Z 1
I= 1 dz = z+c
2 2s
1 2 1
= 2− + 2 +c
2 u u
1 2 1
= 2− 2 + 4 +c
2 x x

4 2
x − 2x + 1
= +c (7)

So, at long last, we have arrived at (A) as the correct choice.
We had to use two substitutions to get to it. The first one was rather
common. But the second one was quite√ tricky. The crucial idea was to
take out a factor u from the radical 2u2 − 2u + 1 in the denominator.
There is a slightly easier way to think of this than the one outlined
above. We rewrite the expression under the radical sign as u2 +(u−1)2.
q taking out the factor u from it, we can rewrite the radical as
1 2
u 1 + (1 − u ) and hence the integral as

1 u−1
I= q du (8)
2 u3 1 + (1 − u1 )2

u−1 1 1
Now we observe that 3
can be rewritten as (1 − ) × 2 . But the
u u u
1 1
second factor 2 is simply the derivative of 1 − w.r.t. u. It follows
u u
1 (u − 1)du
therefore that if we put t = 1 − , then is simply tdt. So,
u Z u3
1 t
with this substitution, I becomes √ dt. This is easy enough
2 1 + t2
to evaluate directly as 1 + t2 . (If one wants, one can try one more
substitution, say y = t2 + 1. But after the tricky substitution we have
used, this one is too straightforward and common.) Converting from t
to u and then from u to x we get the same answer as before.
There does not seem to be any easier way of finding the antideriva-
tive. Even for the methods given above it is difficult to say whether
one could have thought of them without looking at the format of the
answers. Had the answers not been given this problem would have been
quite challenging and a credit of only three points is grossly inadequate
for it. In terms of proportionate time, this means that the question
was meant to be answered in less than two minutes! It is possible that
the intention of the paper-setters was that it should be answered only
by trying the given alternatives one-by-one. Otherwise, instead of ask-
x2 − 1
ing for the indefinite integral √ dx, they could have
x3 2x4 − 2x2 + 1
x2 − 1
asked some definite integral, say √ dx. In that case
x3 2x4 − 2x2 + 1

the answer would have been numerical and it would have been next to
impossible to arrive at it by any short cut.
Q. 9 Suppose f ′′ (x) = −f (x) and g(x) = f ′ (x). Let F (x) = f ( x2 ) +
g( x2 ) . Given that F (5) = 5, F (10) equals
(A) 0 (B) 5
(C) 10 (D) 15

Answer and Comments: (B). Interlinked equations seem to be get-

ting popular with the JEE paper-setters. Last year (2005), there was
a question about two interlinked functional equations. This time we
have a pair of interlinked differential equations. In view of the sec-
ond equation g(x) = f ′ (x), the first equation could have as well been
given as f (x) = −g ′ (x). From this one can, of course, derive that
f ′′ (x) = −f (x). But the paper-setters have spared the candidates by
giving it directly
f ′′ (x) = −f (x) (1)
Apparently, the idea here is to help the candidates. Even though second
order differential equations are not meant to be solved at the JEE
level, this particular equation is very familiar to students as it occurs
in the study of simple harmonic motions in physics. It can be solved
as indicated in Exercise (19.13) and the general solution is
f (x) = A sin x + B cos x (2)
where A and B are arbitrary constants. Since g(x) = f ′ (x), we have
g(x) = A cos x − B sin x (3)
We are not given sufficient data to determine the values of the constants
A and B. But that hardly matters, because no matter what they are,
from (2) and (3) we always have
(f (x))2 + (g(x))2 = A2 + B 2 (4)
for all x. Introducing the notation in the statement of the problem,
this means
F (2x) = A2 + B 2 (5)

for all x. But this means that the function F is identically constant. So
if its value at some point is 5, then this is also the value at all points.
In particular, F (10) = 5.
This is a perfectly legitimate, albeit somewhat sneaky way to
solve the problem which has become possible because of the knowledge
of the general solution of (1). A solution not using (2) is also possible.
Let us multiply both the sides of (1) by 2f ′ (x) . Then we get
2f ′′ (x)f ′ (x) + 2f (x)f ′ (x) = 0 (6)
We recognise the two terms on the L.H.S. as the derivatives of (f ′ (x))2
and (f (x))2 respectively. Keeping in mind that f ′ (x) = g(x), (6) can
be recast as
(g(x))2 + (f (x))2 = 0 (7)
which implies that ((g(x))2 + (f (x))2 ) is identically constant. (As ham-
mered in the remarks about Theorem 3 in Comment No. 9 of Chapter
13, a rigorous proof of this fact is non-trivial and requires the Lagrange
Mean Value Theorem. But in a multiple choice test, what matters is
the result and not whether you can justify it!) So, we have proved (4)
without using (2). The rest of the work remains the same. (We can also
get F ′ (x) ≡ 0 by differentiating F (x) as given and then substituting
from the data.)
Problems where the value of a function at some point is to be
evaluated by showing that the function is identically constant have
appeared in the JEE before. (See for example, the Main Problem of
Chapter 16 or the 1996 JEE problem at the end of Comment No. 7
of Chapter 24.) And, almost invariably, this is done by showing that
the derivative of the function vanishes identically. (For an exception,
see the 1997 JEE problem in Comment No. 4 of Chapter 16.) A
gambler who draws on familiarity with such problems may instinctively
think that the present problem is also of this type and he is right (and
rewarded too by the time saved)!
Q. 10 The axis of a parabola is √along the line y = x. The distance
√ of its
vertex from the origin is 2 and that from its focus is 2 2. If the
vertex and the focus both lie in the first quadrant, the equation of the
parabola is

(A) (x + y)2 = (x − y − 2) (B) (x − y)2 = (x + y − 2)
(C) (x − y)2 = 4(x + y − 2) (D) (x − y)2 = 8(x + y − 2)

Answer and Comments: (None). The focus and the vertex of any
parabola always lie on its axis. Call the focus as F and the vertex as V .
In the present problem, the axis is given as the line y = x and therefore
v = (a, a) and F = (b, b) for some real numbers a, b which are positive
as both V and F lie in the first √quadrant. We are also given that the
distance of V from the origin is 2. This determines a as 1 and hence
V√= (1, 1). We are further given that the distance between V and F is
2 2. This means (b − 1)2 + (b − 1)2 = 8, which implies b = 1 ± 2 = 3
since the − sign is excluded by the positivity of b.
So, we have determined the focus F as (3, 3). If we can now
determine the directrix, say L, of the parabola then we shall get the
equation of the parabola. The directrix is always perpendicular to the
axis, and hence in the present problem its slope is −1. To determine it,
we need to know any one point, say G on it. The best choice is to take
G as the point of intersection of the directrix and the axis. This means
G is of the form (c, c) for some c ∈ IR. Further, the vertex always lies
midway between the focus F and this point G. As we already know the
focus as (3, 3) and the vertex as (1, 1), c is determined by the equation
= 1, i.e. c = −1. Hence the point G is (−1, −1). As we already
know the slope of the directrix L to be −1, its equation comes out to

x + y = −2 (1)

Having known the directrix L and the focus F , we get the equa-
tion of the parabola straight from its definition, viz. the locus of a
point which is equidistant from L and F . The distance of a point
|x + y + 2|
P = (x, y) from the line (1) is √ while its distance from F is
q 2
(x − 3)2 + (y − 3)2 . Hence the equation of the parabola is

(x + y + 2)2 = 2((x − 3)2 + (y − 3)2 ) (2)

which, upon expansion and simplification, becomes

(x − y)2 = 16(x + y − 2) (3)

which does not match with any of the given alternatives. Had the focus
been at (2, 2) (instead of at (3, 3)), the point G would have been the
origin (instead of (−1, −1)) and the equation of the directrix L would
have been x + y = 0. The equation of the parabola would then have
been (x+ y)2 = 2((x−2)2 + (y −2)2 ) and after simplification this would
have matched the answer (D).
Apparently, the confusion arises because of the phrase ‘and that
from its focus’ in the statement of the problem. As the problem reads,
it means the distance of V from F rather than the distance of the origin
from F . If the latter meaning was intended then √ instead of saying that
‘the distance of its vertex from the origin is 2 ’, the correct wording√
should have been ‘the distance of the origin from its vertex is 2 ’.
Mathematically, the distance is symmetric and so it does not matter
whether you say ‘the distance of A from B’ or ‘the distance of B from
A’. But the way the word ‘that’ is used in English, the interpretation
of the second part (involving the focus) changes drastically. It is inter-
esting to note that if instead of saying ‘that from its focus’, the problem
had said ‘that of its focus’ then it√would have meant that the distance
of the focus from the origin is 2 2. This would have fixed the focus
at (2, 2) and, as shown above, (D) would have been the answer. How
the change of a single word can change the problem! This needless
ambiguity could have been avoided by simply giving the vertex and
the focus directly. It would be a pity if a student loses 4 marks (not
to mention some precious time) just because of poor grammar whether
on his own part or on the part of the paper-setters! (It may, of course,
very well be that the mistake is not in the original question paper, but
in its memorised version.)
In fact, even from a strictly mathematical point of view, it is
difficult to see what has been achieved by giving the vertex and the
focus of the parabola in such a twisted manner. The problem is about
parabolas and its central idea is that the vertex and the focus lie on the
axis and further that the vertex is equidistant from the directrix and
the focus. Even if the vertex and the focus had been given explicitly
as (1, 1) and (3, 3) (or (2, 2) depending on the intention), the major
work needed would have been the same. In fact, then the quantum of
the work would have been fair for a three point question. The work
needed to identify the vertex and the focus first has little to do with

the subsequent work and therefore represents only an additional burden
which serves little purpose as far as testing the main idea is concerned.
ex ,


Q. 11 If f (x) =  2 − e , 1 < x ≤ 2 and g(x) = f (t)dt, then g(x) has

x − e, 2<x≤3

(A) a local maximum at x = 1 + ln 2 and a local minimum at x = e

(B) a local maximum at x = 1 and a local minimum at x = 2
(C) no local maxima
(D) no local minima

Answer and Comments: (A). To determine the maxima/minima of

g(x) we first identify its critical points, i.e. points where g ′(x) either
vanishes or fails to exist. As the function g(x) is defined by an integral,
its derivative g ′(x) is given by the second form of the Fundamental
Theorem of Calculus (Theorem 2 in Comment No. 11 of Chapter 17).
It is simply f (x). Thus we have

ex ,

 0≤x≤1
g ′(x) = 2 − ex−1 , 1 < x ≤ 2 (1)
x − e, 2<x≤3

As the exponential function is always positive, the first line on the

R.H.S. of (1) shows that g ′(x) has no zero in [0, 1]. Since 2 < e < 3,
the third line shows that g ′ (x) has one zero, viz. e in (2, 3]. The
zeros in (1, 2] are precisely the solutions of the equation ex−1 = 2, or
equivalently x − 1 = ln 2. There is only one solution, viz. x = 1 + ln 2.
It is important to note that this point indeed lies in the interval (1, 2]
because 0 < ln 2 < 1 (since 1 < 2 < e).
Summing up, g ′(x) vanishes at 1 + ln 2 and e. So these are
among the candidates for the local maxima/minima of g(x). Further,
differentiating (1), we get

ex ,

 0≤x<1
g (x) = −ex−1 , 1 < x < 2 (2)
1, 2<x≤3

In particular, we have g ′′ (1 + ln 2) = −eln 2 = −2 < 0, which means
that g(x) has a local maximum at x = 1 + ln 2. Similarly, g ′ (e) = 1 > 0
whence there is a local minimum at x = e. So the statement (A) is
The solution is over because only one option is supposed to be
correct. But in case the question had been designed so that one or
more options are correct, then we would have to check if (B) holds.
(We already know that (C) and (D) are false.) Since the checking of
(B) involves some interesting ideas, we present it here, even though it
is not a part of the solution as the problem now stands.
In addition to the points where g ′(x) vanishes, we also have to look
for those points where g ′(x) fails to exist. From (1), it may appear that
g ′(x) exists at all points in [0, 3]. But it is not quite so. To explain
this apparent anomaly, we need to look at the second fundamental
theorem of calculus more carefully. In its usual form it says that if
f (x) is continuous on [a, b] and g(x) is defined by g(x) = f (t)dt for

a ≤ x ≤ b, then g (x) exists and equals f (x) for all x ∈ [a, b]. Here, if
x equals either of the two end-points a or b, then by continuity of f we
mean only the appropriate left or right continuity. The same holds for
differentiability of g(x) at the end-points.
What happens if f (x) has a removable discontinuity at a? This
means the right handed limit lim+ f (x) exists but does not equal f (a).
As long as f (x) is bounded on [a, b], a finite number of such discon-
tinuities does not affect the integral. So we can still define g(x) as
Rx ′
f (t)dt for x ∈ [a, b]. But now g+ (a), i.e. the right handed derivative
of g(x) at a will equal lim+ f (x) and not f (a). Similarly, if there is

a removable discontinuity at b, then the left handed derivative g− (b)
equals lim− f (x) and not g(b).
The same considerations apply if f (x) has a discontinuity at some
intermediate point, say c ∈ (a, b). As long as the two limits lim− f (x)
and lim+ f (x) exist, there is no difficulty in defining the integral of f
over [a, b] and hence the function g : [a, b] −→ IR by g(x) = f (t)dt.
But now we can no longer say that g(x) is differentiable at c. All we


can say is that g(x) is right differentiable at c with g+ (c) = lim+ f (x)

and similarly, it is left differentiable at c with g− (c) = lim− f (x).
Although these fine considerations are usually omitted at the
JEE level, it is precisely this hair splitting that is needed in the present
problem. Here the integrand f (x) given in the statement of the problem
has two possible discontinuities in its domain [0, 3], one at x = 1 and
the other at x = 2. At x = 1, the left and right handed limits of f (x)
are e and 1 respectively. So at x = 1, Equation (1) is not correct as
stated but instead has to be interpreted to say that

g− (1) = e (3)

and g+ (1) = 1 (4)

Similarly, at x = 2, the left and right handed limits of f (x) are 2 − e

and 2 − e. This makes f continuous at this point and so g ′ (x) = 2 − e.
In other words, x = 2 is not a critical point of g(x). So there is neither
a local maximum nor a local minimum at x = 2. This makes (B) false
even without checking what happens at x = 1. Nevertheless we go into
it from an academic point of view.
Since 1 is a point where even the first derivative of g(x) does
not exist, the question of the second derivative does not arise. So, the
popular and handy ‘second derivative test’ is useless in determining
whether there is a local maximum of g(x) at x = 1. Instead we have to
go by something more basic, and far more elementary. If c is an interior
point of the domain of a (not necessarily differentiable) function g(x),
then g will have a local maximum at c if g changes its behaviour from
increasing to decreasing as x passes over the point c from the left to
right. In other words, if there is some small neighbourhood of c, say
(c − δ, c + δ) such that g(x) is increasing on (c − δ, c) and decreasing on
(c, c + δ), then g(x) has a local maximum at c. In particular this is the
case if g ′ (x) > 0 for all x ∈ (c − δ, c) and g ′ (x) < 0 for all x ∈ (c, c + δ).
Note that g ′(c) is not involved here at all, much less g ′′(c). An entirely
analogous criterion holds for g(x) to have a local minimum at c.
In the present problem, from (1) we see that g(x) is increasing on
the interval (0, 1) and also increasing on (1, 1 + δ) if δ is positive and
sufficiently small (specifically, if 0 < δ < ln 2). So, g(x) is increasing

on both the sides of 1. Hence there is neither a local maximum nor a
local minimum at x = 1.
In fact, this simple criterion, based on a change of the increas-
ing/decreasing behaviour could have been applied even without using
the fundamental theorem of calculus to find g ′ (x). The basic idea is
simply that if the integrand is positive throughout an interval, then the
integral increases as the interval gets larger, while if it is negative then
the integral decreases. More specifically, suppose g(x) = f (t)dt and
[c, d] is an interval contained in the domain of f (x). If f (x) > 0 for all
x in the open interval (c, d), then the function g(x) is monotonically
increasing (in fact, strictly monotonically increasing) on the interval
[c, d]. This follows by writing g(x + h) − g(x) as xx+h f (t)dt and noting

that for h > 0, the integral is positive as the integrand is positive on

(x, x + h). Here we are using only very elementary properties of def-
inite integrals which can be derived directly from their definitions as
limits of Riemann sums. A deep result like the fundamental theorem
of calculus is nowhere needed.
This simple-minded observation works wonders in the present prob-
lem. We simply keep track of where the integrand f (x) is positive and
where it is negative. We use the same reasoning as we applied above to
determine the sign of g ′ (x). But now we are applying it directly to f (x)
as given in the statement of the problem. The answer is, of course the
same, viz. that f (x) is positive on [0, 1 + ln 2), negative on (1 + ln 2, e)
and then positive again on (e, 3]. So, by our simple criterion, g(x) has
a local maximum at x = 1 + ln 2, a local minimum at e and no other
local maxima or minima. So, once again (A) is the only true answer.
When done by the conventional method (based on derivatives),
the problem requires certain subtleties of the fundamental theorem of
calculus to check whether (B) is also a correct answer. The problem
would have been far more interesting if, on the interval (1, 2], f (x)
were given as ex−1 − 2 instead of 2 − ex−1 . In that case the function
f (x) would have changed its sign at all the four points 1, 1 + ln 2, 2
and e and both (A) and (B) would have been correct. In fact such a
problem would have been commendable as an eye-opener to those who
have acquired the dirty habit of indiscriminately applying derivatives to
tackle any problems of local maxima and minima. This habit becomes

so dominating that many people think that to say that a function is
strictly increasing over an interval means that its derivative is positive
at every point of that interval, totally forgetting that the concept of
monotonicity is in terms of comparison of functional values and that
derivatives are only a convenient tool which works often but not al-
ways. It is probably a confusion like this which resulted in an incorrect
problem in the Screening Paper of JEE 2004 Mathematics. (See Q. 2
in the author’s commentary on the same.)

Q. 12 Let ~a = î + 2ĵ + k̂, ~b = î − ĵ + k̂ and ~c = î + ĵ − k̂. A vector in the

plane of ~a and ~b whose projection on ~c is √ , is
(A) 4î − ĵ + 4k̂ (B) 3î + ĵ − 3k̂
(C) 2î + ĵ + 2k̂ (D) 4î + ĵ − 4k̂

Answer and Solution: (C). Although the methods needed to solve

this problem are straightforward, what makes it unusual is that there is
no unique answer to it. The given requirements on the desired vector,
say ~v , do not determine it uniquely. Let P be the plane spanned by the
vectors ~a and ~b. Resolve ~c along and perpendicular to P , i.e. write ~c as
~u + w~ where ~u is in P and w ~ is perpendicular to P . As ~v is given to lie
in P , its projection on ~c is the same as its projection on ~u. Resolve ~v
as ~x + ~y , where ~x is parallel to ~u and ~y is perpendicular to ~u. Then the
projection of ~v on ~u depends only on ~x and not on ~y . By changing the
magnitude of ~y , we get infinitely many vectors which satisfy the same
conditions as ~v .
That is why the problem only specifies ‘a’ and not ‘the’ vector which
satisfies the given conditions. The right way to solve the problem is to
identify the set of all vectors that satisfy its conditions and then see
which of the given alternatives belongs to it.
Once this point is understood, the problem itself is simple. Let
~v = v1 î + v2 ĵ + v3 k̂ be a vector which satisfies the given conditions.
The first condition implies that ~v is perpendicular to the cross product
~a × ~b. By a direct computation,

î ĵ k̂

~a × b = 1 2 1 = 3î − 3k̂


1 −1 1

As ~v is perpendicular to ~a × ~b, we get

v1 = v3 (2)
The second requirement on ~v is that its projection along ~c is √ . Since
1 î + ĵ − k̂
~c = î + ĵ − k̂, a unit vector along ~c is √ ~c = √ . Therefore this
3 3
~v · ~c 1
requirement gives √ = √ , i.e. v1 + v2 − v3 = 1. In view of (2), this
3 3

v2 = 1 (3)

(2) and (3) give the general form of a vector which satisfies the given
conditions. Out of the given choices, (C) is the only one where they
are satisfied.
A slightly different approach is to start by taking ~v as a linear
combination of the vectors ~a and ~b, say,

~v = α~a + β~b
= α(î + 2ĵ + k̂) + β(î − ĵ + k̂)
= (α + β)î + (2α − β)ĵ + (α + β)k̂ (4)

where α, β are some scalars. As before, the second requirement becomes

[(α + β)î + (2α − β)ĵ + (α + β)k̂] · (î + ĵ − k̂) = 1 (5)


2α − β = 1 (6)

Putting (6) into (4) we get

~v = (3α − 1)î + ĵ + (3α − 1)k̂ (7)

Here α can have any real value. This gives us the same set of vectors
as before and out of the given options (C) is the only one where the
given vector belongs to this set.

One or more of the given answers is/are correct.

Q. 13 The equation(s) of the common tangent(s) to the parabolas y = x2 and

y = −(x − 2)2 is/are
(A) y = 4(x − 1) (B) y = 0
(C) y = −4(x − 1) (D) y = −30x − 50

Answer and Comments: (A), (B). This is a straightforward prob-

lem. It can be done by various methods which differ more in presen-
tations than in substance. The best method is to take a typical line
y = mx+c and determine for which value(s) of m and c it touches both
the parabolas. (In doing this there is a danger that we may miss com-
mon tangents whose equations cannot be written down in this form.
These are the vertical lines, i.e. lines of the form x = a for some con-
stant a. But in the present problem, the first parabola, viz. y = x2
is such a standard figure that even without drawing it, it is clear that
it has no vertical tangents. So, nothing is missed. But this point
represents an inherent limitation of an objective type test. A careful
student will spend some time to rule out this possibility, while to an
unscrupulous student, it may simply not occur! And there is no way to
distinguish because no reasoning is to be given. In short, in a question
like this, ignorance is bliss.)
So suppose y = mx+c is a line in the plane. Its points of intersection
with y = x2 correspond to the roots of the quadratic equation

x2 = mx + c (1)

the line will be a tangent to the parabola if and only if these roots
coincide, i.e. the discriminant vanishes. This gives

m2 + 4c = 0 (2)

as the condition for tangency. (For a parabola in the standard form,

viz. y 2 = 4ax, the condition for tangency of a line y = mx + c is

a standard formula. If you can modify it correctly, you can apply it
here and save some time. But this is prone to error. At any rate, it is
always a good idea to know the underlying reasoning, which requires
no modification whether the parabola is in the standard form or not.)
By an analogous reasoning, y = mx + c touches the parabola
y = −(x − 2)2 if and only if the quadratic (x − 2)2 + mx + c = 0 has a
vanishing discriminant, i.e.

(m − 4)2 − 4(c + 4) = 0 (3)

Adding (2) and (3), we get 2m2 − 8m = 0, which means m = 0 or

m = 4. We can now find the corresponding values of c from either (2)
or (3). But that is not necessary. In the given options (A) and (B)
are the only ones where the lines given have slopes 0 or 4. So we mark
them without further ado. (Such cheap short cuts are academically
insignificant. But in an examination they can save you precious time.)
x2 y 2
Q. 14 If a hyperbola passes through the focus of the ellipse + = 1 and
25 16
its transverse and conjugate axes coincide with the major and minor
axes of the ellipse, and the product of the eccentricities is 1, then
x2 y2
(A) the equation of the hyperbola is − =1
92 162
x y
(B) the equation of the hyperbola is − =1
9 25
(C) focus of the hyperbola is (5,√0)
(D) focus of the hyperbola is (5 3, 0)

Answer and Comments: (A), (C). Looks more like a problem de-
signed to test the knowledge of the vocabulary about conics! Less
pretentiously, the second condition simply means that the equation of
the hyperbola can be taken to be in the standard form

x2 y 2
− 2 =1 (1)
a2 b
So the hyperbola will be determined as soon as we know the values
of a and b (which can be taken to be positive). For this we need two
equations in these two unknowns. And these are provided by the other

parts of the data. The condition about passing through the focus of
the ellipse is a little faultily expressed. Every ellipse has two foci and
so the language ‘the focus’ is incorrect. Fortunately, in the present
case both the foci of the ellipse are symmetric about the y-axis. As the
hyperbola is also symmetric about the y-axis, once it passes through
either focus, it also passes through the other.
Now, coming to the solution, the eccentricity of the given ellipse is
√ 3
given by 5 1 − e2 = 4 which determines e as . Hence the foci of the
ellipse are at (±3, 0). As both these points satisfy (1), we have
=1 (2)
which gives a = 3.
The third condition in the data determines the eccentricity, say
e′ of the hyperbola as e′ = . Since the eccentricity of the hyperbola
√ 3 √
a2 + b2 5 b2 + 9
(1) is and a = 3, we get = which determines b as
a 3 3
√ 5
25 − 9 = 4. Finally, since the foci of (1) lie at (±ae′ , 0), from e′ =
and a = 3 we get that the foci of the hyperbola lie at (±5, 0).

Q. 15 Internal bisector of 6 A of a triangle ABC meets the side BC at D. A

line drawn through D and perpendicular to AD intersects the side AC
at E and the side AB at F . If a, b, c represent the sides of ∆ABC,
(A) AE is H.M. of b and c (B) AD = cos A2
4bc A
(C) EF = sin 2 (D) the triangle AEF is isosceles

Answer and Comments: (All). Since AD is the altitude as well

as the internal angle bisector through A of the triangle AEF (D) is
obvious. The remaining three statements are based on the formula for
the length of an angle bisector of a triangle. (B) is this very formula.
Although not as standard as others, it is a fairly well-known formula,

see e.g. p. 246 of Trigonom-
etry by S. L. Loney. Once
we know AD we get AE A /2
immediately as AD sec 2 = c
. This makes (A) b
b+c B
true. Finally, we also F E
have ED = AD tan A2 = D
2bc A
sin 2 . Since we al- C
ready know that the trian-
gle AEF is isosceles, AD is
also a median. Therefore
EF = 2ED. Hence (C) is
also true.
All except one of the alternatives are easy consequences of a rela-
tively less known formula in trigonometry. Those who have memorised
this formula get a rather substantial advantage in terms of time saved.
In this respect, the problem resembles Q. 3 in Section A.

Q. 16 If f (x) = min{1, x2 , x3 }, then

(A) f (x) is continuous at every x ∈ IR
(B) f ′ (x) > 0 for every x > 1
(C) f (x) is continuous but not differentiable for every x ∈ IR.
(D) f (x) is not differentiable for two values of x

Answer and Comments: (A), (C). A common problem about test-

ing continuity and differentiability of a given function f (x). The only
difference is that f (x) is given in a somewhat unusual manner, viz. as
the smallest of the three numbers 1, x2 and x3 . So the first task is
to decide which of these three expressions equal f (x) in which of the
intervals of the real line.
There are three functions here, the constant function 1, the function
x and the function x3 . They are all continuous. Now, if f1 (x), f2 (x)

are continuous functions, then to decide which is greater for which x,

we identify all points where the two are equal. Equivalently, we find
the zeros of the difference function f1 (x) − f2 (x). In between any two
consecutive zeros, we shall have that either f1 (x) > f2 (x) for all x or

else f1 (x) < f2 (x) for all x. (This conclusion is intuitively obvious and
we often use it as a preliminary step in finding things like the area
between the graphs of the two functions. But a rigorous proof requires
the Intermediate Value Property of continuous functions applied to the
function f1 − f2 .)
Now, coming to the present problem, we are fortunate that all the
three functions 1, x2 and x3 agree at x = 1. The first two also agree
at x = −1 while the last two also agree at x = 0. Therefore −1, 0
and 1 are the points we have to be wary about. The graphs of these
functions are very well-known. But even without them, it is easy to see
that among these three functions, x3 is the smallest for x < 0, because
it is negative while the other two are positive. Also, for x > 1, both
x2 and x3 exceed 1 and so the minimum of the three functions is 1.
For x ∈ (0, 1), we have x3 < x2 < 1 and so the minimum of the three
functions is x3 .
We are now in a position to cast the given function f (x) in a more
conventional form, viz.
x3 if x ≤ 1
f (x) = (1)
1 if x > 1

Evidently, f (x) is differentiable (and hence continuous) at every

x 6= 1. Also, the right and the left handed limits of f (x) equal 1 each.
So, it is also continuous at 1. For differentiability at 1, we need to
check the left and right handed derivatives of f (x) at x = 1. These are
respectively, 3 and 0. (The latter is obvious. The simplest justification
for the former is that the function x2 is continuously differentiable on
(−∞, 1]. So its (left handed) derivative is the limit of its derivative 3x2
as x → 1−1 .)
Thus we see that f (x) is continuous everywhere and differentiable
everywhere except at one point, viz. x = 1. This renders (A) and (C)
true and (D) false. As for (B), f (x) is a constant for x > 1 and so its
derivative vanishes identically.
Once the function f (x) is identified in the form (1), the problem
is very simple. Coming up with (1) requires an elementary knowledge
of inequalities involving powers. The problem is a good combination of
two elementary ideas and requires virtually no computation.

Were the problem only about continuity of f (x), we could have
bypassed (1). The minimum (and also the maximum) of two continu-
ous functions is continuous. This follows by writing min{f1 (x), f2 (x)}
and max{f1 (x), f2 (x)} as 21 (f1 (x) + f2 (x) ∓ |f1 (x) − f2 (x)|) respectively
and using the continuity of the absolute value function. By induction,
the result holds for the max/min of any finite number of continuous
functions. Unfortunately, for differentiability there is no such short

Q. 17 f (x) is a cubic polynomial which has a local maximum at x = −1. If

f (2) = 18, f (1) = −1 and f ′ (x) has local maximum at x = 0, then

(A) the distance between (−1, 2) √ and (a, f (a)), where x = a is the
point of local minimum is 2 5

(B) f (x) is increasing for x ∈ [1, 2 5]
(C) f (x) has local minimum at x = 1
(D) f (0) = 5

Answer and Comments: (B), (C). This problem is strikingly similar

to Problem 12 of the Main paper of 2005 JEE Mathematics (see the
author’s commentary on the same), which, in turn, is of the same spirit
as the Main Problem of Chapter 15 or Exercise (17.23). In all these
problems, we are given some data about a cubic polynomial and the
starting point is by converting the data to a system of equations whose
solution will determine the cubic uniquely.
So, in the present problem, we let f (x) = ax3 + bx2 + cx + d.
(This choice of notation is a little dangerous because the same symbol
a is used in the statement of the problem to denote something else,
viz. the point where f (x) has a local minimum. But no confusion need
arise because we are first going to determine a, b, c, d from the given
conditions and there is no harm if later on the same symbols are used
for something else. Still, those who want to play it safe, may take f (x)
as px3 +qx2 +rx+s or, more clumsily, as a0 x3 +a1 x2 +a2 x+a3 .) We need
four equations to determined these four unknowns a, b, c, d. The four
conditions in the data give these equations. But instead of writing them
mechanically one after another, it is better to see if some short cuts are

possible by using the pieces of the data in a more clever order. The last
piece of data implies that f ′′ (x) vanishes at x = 0. Since f (x) has degree
3, f ′′ (x) is of degree 1. So an equation involving it is much easier to deal
with. Specifically, we have f ′ (x) = 3ax2 +2bx+c and f ′′ (x) = 6ax+2b.
So f ′′ (0) = 0 gives us b = 0. This makes f (x) = ax3 + cx + d and
simplifies the other three equations we are going to get from the other
three pieces of the data. Now that f ′ (x) = 3ax2 + c, the first piece
implies that f ′ (−1) = 0 i.e. c = −3a. So, now we take f (x) as
ax3 − 3ax + d. We now have only two unknowns, viz. a and d. To get
their values, we use the remaining two pieces of data, viz. f (2) = 18
and f (1) = −1, which translate, respectively, into
2a + d = 18
and − 2a + d = −1 (1)
which can be solved by inspection to get a = 19
and d = 34
. Hence we
57 1 3
also get c = −3a = − 4 . Thus f (x) = 4 (19x − 57x + 34).
Now that we have got hold of f (x), we can answer any questions
about it one-by-one. But once again, it is better to begin with the more
direct ones first. (D) requires only the constant term in f (x), which is
and not 5. So, (D) is false. Both (B) and (C) require the derivative
f (x) = 57

(x2 − 1) which vanishes at x = ±1. As we are already given
that there is a local maximum at x = −1, it follows from the general
properties (given in Comment No. 13 of Chapter 13) of a cubic that
there is a local minimum at the other critical point, viz. x = 1. Of
course, we can also verify this directly by computing f ′′ (1) as 57

is positive. So, (C) is true. Again, f (x) > 0 for all x > 1 and √ so
f is strictly increasing on [1, ∞) which includes the interval [1, 2 5].
Hence (B) is also true. Finally, for (A), we already know that a = 1.
(We remark again that this a is different from the one that appeared
earlier and in particular in (1) above. We now know the value of the
former a, viz. a = 19 4
. The present a is the point where f (x) has a
local minimum.) Hence (a, f (a)) = (1, f (1)) = (1, −1). The distance
of √this point√from (−1, 2) (which
√ is not a point on the graph of f (x))
is 4 + 9 = 13 and not 2 5. So, (A) is false.
The computations involved are simple but prone to errors. Con-
ceptually, there is nothing exciting in the problem. The problem would
have been more interesting had the data been insufficient to identify

f (x) uniquely, but nevertheless sufficient to answer the given ques-
tions. As it stands, it is straightforward almost to the point
√ of being
a drudgery. The significance of the particular number 2 5 which ap-
pears twice in the problem is far from clear. Maybe it is a random
figure inserted only to test if a candidate can save himself from getting
confused. Part (A) merely involves finding the distance between two
points in the plane. It is not clear what purpose it serves, except to
make life miserable for a student by increasing his labour and chances
of numerical errors. The language in Part (B) is a little faulty too. A
function is said to be increasing (or decreasing) over an interval, rather
that at a point of the interval.

Q. 18 A is a 3×3 matrix and ~u is a column vector. If A~u and ~u are orthogonal

for all real ~u, then the matrix A is
(A) singular (B) non-singular
(C) symmetric (D) skew-symmetric

Answer and Comments: (A), (D). Let

   
a1 b1 c1 x
A =  a2 b2 c2  and ~u =  y  (1)
   

a3 b3 c3 z

Then by a direct computation,

 
a1 x + b1 y + c1 z
A~u =  a2 x + b2 y + c2 z  (2)
 

a3 x + b3 y + c3 z

Orthogonality of A~u and ~u means that their dot product vanishes, i.e.

x(a1 x + b1 y + c1 z) + y(a2 x + b2 y + c2 z) + z(a3 x + b3 y + c3 z) = 0

or, after simplification,

a1 x2 + b2 y 2 + c3 z 2 + (a2 + b1 )xy + (c2 + b3 )yz + (c1 + a3 )zx = 0 (3)

If this is to hold for all x, y, z, the coefficient of every term must vanish.
So a1 = b2 = c3 = 0 and a2 = −b1 , c2 = −b3 , a3 = −c1 . Therefore the

matrix A looks like
 
0 b1 c1
 −b1
A= 0 c2  (4)
−c1 −c2 0

It is immediate that A is skew-symmetric. Also, its determinant comes

out to be 0. So it is singular too. (More generally, one can show
that for odd n, every skew-symmetric matrix of order n is singular.
For, if A is such a matrix, then by definition, At , i.e. the transpose
of A, equals −A. Since every matrix has the same determinant as its
transpose, we get that A and −A have the same determinant. But
| − A| = (−1)n |A| = −|A| as n is odd. So, |A| = −|A|, which means
|A| = 0.)
If one wants, the first part (that of showing the skew-symmetry
of A) can also be done more elegantly. The orthogonality of A~u with
~u can be paraphrased to say that ~u t A~u = 0. Taking transposes of
both the sides, ~u t At~u = 0 and hence ~u t P ~u = 0 for all ~u, where
P = A + At . Symmetry of P then forces it to vanish. But even without
these elegant solutions, the problem is a good one as the computations
do not dominate the concepts.
Note that the L.H.S. of (3) is simply ~u tA~u where A and ~u are as in
(1). It is a homogeneous polynomial of degree 2 in the three variables
x, y, z. Such expressions are called quadratic forms. They have many
applications, including classification of conics. And matrices, especially
the symmetric ones, play an invaluable role in their analysis.

Q. 19 A tangent drawn to the curve y = f (x) at P (x, y) cuts the x-axis and
the y-axis at A and B respectively so that BP : AP = 3 : 1. Given
f (1) = 1,

(A) the equation of the curve is x − 3y = 0
(B) normal at (1, 1) is x + 3y = 4
(C) the curve passes through (2, 1/8)
(D) the equation of the curve is x + 3y = 0

Answer and Solution: (C), (D). Another problem which is strikingly
similar to a JEE problem in the past, specifically the 1998 JEE problem,
solved in Comment No. 16 of Chapter 19. The only difference in fact,
is that in the earlier problem it was given that BP = AP while in the
present problem we have BP : AP = 3 : 1.
So, once again we take P = (x0 , y0) as a typical point on the given
curve and let m be the slope of the tangent to the curve at this point.
Then the equation of the tangent to the curve at P is

y = y0 + m(x − x0 ) (1)

The points A and B come out to be respectively

A = (x0 − , 0) and B = (0, y0 − mx0 ) (2)
We are given that the point P divides the segment AB in the ratio
1 : 3. Using the section formula and (2) this gives
3(x0 − m
) y0 − mx0
x0 = and y0 = (3)
4 4
(These two equations are exactly the same. This is a consequence of
the collinearity of A, P and B. In fact, with a little foresight, we could
have stopped after writing either one of them down as we already know
that the other one can convey no new information.)
Using either of these two equations we get

mx0 + 3y0 = 0 (4)

We now replace x0 , y0 , m (which were introduced to avoid confusion in

an equation like (1)), by x, y and respectively and get
x + 3y = 0 (5)
as the equation of the curve. So (D) is true and (A) is false. (Actually,
this is a differential equation and represents a one-parameter family
of curves, of which the given curve is one member. So the language
‘equation of the curve’ used in the problem is slightly misleading.)

The next task is to solve (5). This is very easy. Rewriting it in the
differential form as xdy = −3ydx and then in the separate variables
dy dx
form as = −3 , the general solution is ln y = − ln(x3 ) + c or
y x
equivalently, yx3 = k where c and k are some constants. As the curve
passes through (1, 1) we get k = 1. Therefore the equation of the curve
y= (6)

Now that we know the curve completely, we can answer any

questions about it. It is obvious that it passes through the point (2, 81 ).
So, (C) holds. Finally, to find the normal at (1, 1), we already know
from (5) that the slope of the tangent at (1, 1) is −3. So the slope
of the normal is 13 . But the line given in (B) has slope − 13 . So, even
without finding the normal at (1, 1), we know that (B) is false.
The essential part of the problem is over as soon as (6) is obtained.
Determining whether the statements (B) and (C) are true is an addi-
tional drudgery which has absolutely nothing to do with the crux of
the problem. It also makes it possible that a candidate who gets the
heart of the problem correctly, later makes a silly slip in working over
these appendages and thereby gets −1 mark instead of 5 which he fully
deserves. In a keenly competitive examination like JEE, losing 6 points
may translate into the difference between getting in and not getting in.
In a conventional examination, the problem would most likely have
asked the candidates only to find the equation of the curve. But in a
multiple choice test, it is dangerous to do so, because if (6) is given
as one of the answers, one can simply verify it rather than obtain it
honestly by forming and solving a differential equation. An excellent
example of how a particular format of an examination gives undue
importance to some peripheral things.
For the analogous 1998 JEE problem mentioned above, the work
was essentially the same (and, in fact, a little simpler) and the equation
of the curve came out to be y = 1/x, because even the initial condition
(viz. f (1) = 1) is the same for both the problems! It is indeed shocking
that essentially the same question is repeated. The explanation perhaps
is that although differential equations form an extremely vast area of

mathematics, only a very tiny fragment of it is included at the JEE
level and so there is not much scope to come up with qualitatively new
problems every year.
There is also the time factor. In the 1998 JEE, the question carried
8 points in a 200 point test to be done in 3 hours. Proportionately, this
meant a little over 7 minutes for the problem. The present problem
has 5 points in a 184 marks test for 2 hours. So, proportionately, you
get less than half the time as in 1998. And you have to do more work.
Working with the midpoint is not as time consuming as working with
a point which divides a segment in some other proportion. Moreover,
even after getting the equation of the curve you have to tackle the
useless appendages as pointed out earlier. It is no consolation that in
1998 you had to show the work while now you need not. This argument
would hold some water in the case of problems like Q. 1 or Q. 7 above.
But in a highly computational problem like the present, one still has to
do the work in the rough, whether one displays it later or not. Maybe
the intention behind asking a familiar type problem was to make up
for this severe reduction in the time allowed.

Q. 20 Let A ~ be a vector parallel to the line of intersection of the planes P1

and P2 through the origin. P1 is parallel to the vectors 2ĵ + 3k̂ and
4ĵ − 3k̂ and P2 is parallel to ĵ − k̂ and 3î + 3ĵ. Then the angle between
the vectors A ~ and 2î + ĵ − 2k̂ is
π π
(A) (B)
2 4
π 3π
(C) (D)
6 4

Answer and Comments: (B), (D). This is a straightforward problem

about the cross and the dot products of vectors. We first need to
determine the direction of the vector A. ~ Let N ~ 1 and N ~2 be vectors
normal to the planes P1 and P2 . As A ~ is parallel to line of intersection
of P1 and P2 , it is perpendicular to both N1 and N2 . Therefore, it is
~1 × N
parallel to N ~ 2 . This does not determine A ~ uniquely. But that is
not needed either, because all we want to find is the angle between A ~
and the vector 2î + ĵ − 2k̂.
So, we must first determine N ~1 and N ~ 2 . This is easy because we

are given two (linearly independent) vectors in each plane and forming
~1, N
the cross products of these pairs, we get N ~2 . Thus,

î ĵ k̂

N = (2ĵ + 3k̂) × (4ĵ − 3k̂) = 0 2 3

= −18î (1)

0 4 −3

î ĵ k̂

and N = (ĵ − k̂) × (3î + 3ĵ) = 0 1 −1 = 3î − 3ĵ − 3k̂ (2)

3 3 0

(A slight short-cut to (1) is possible. Note that we are interested

~ 1 which is a normal to the plane P1 . Since P1
only in the direction of N
contains the vectors 2ĵ + 3k̂ and 4ĵ − 3k̂, neither of which has any
component parallel to î, it is obvious that this plane P1 is simply the
yz-plane. Therefore we may as well take N ~1 = î without having to
do any computations. Such a clever thinking is not possible for N ~2 .
And even when it is possible, the time to recognise its applicability is
probably the same as the time to get the result by computation. But
it does serve some purpose in confirming an answer.)
As noted above, we may take A ~=N ~1 × N
~2 . This gives,

î ĵ k̂

A = (−18î) × (3î − 3ĵ − 3k̂) = −18 0


= −54ĵ + 54k̂ (3)

3 −3 −3

~ and the vector 2î + ĵ − 2k̂.

Let θ be either of the two angles between A
~ · (2î + ĵ − 2k̂)
cos θ = ±
~ î + ĵ − 2k̂)|
(−54ĵ + 54k̂) · (2î + ĵ − 2k̂)
= ±
|(−54ĵ + 54k̂)||(2î + ĵ − 2k̂)|
= ± √ √
54 2 × 9
= ±√ (4)

π 3π
which gives θ = or . So, (B) and (D) are correct.
4 4
Note that the vector A ~ is the cross product N ~1 × N~2 , where each
of the vectors N~1 and N ~ 2 is itself a cross product of two vectors. So A~
is a vector of the form
A~ = (~a × ~b) × (~c × d)
~ (5)
where ~a = 2ĵ + 3k̂, ~b = 4ĵ − 3k̂, ~c = ĵ − k̂ and d~ = 3î + 3ĵ. Now,
there is an identity for the vector appearing in the R.H.S. of (5), which
expresses it in terms of certain scalar triple products (or ‘box’ products)
of vectors, viz.
(~a × ~b) × (~c × d)
~ = (~a ~c d)
~ ~b − (~b ~c d)~
~a (6)
(For a proof as well as an application, see Comment No. 12 and also
Comment No. 18 of Chapter 21.) So we could as well get (3) directly
using (6), bypassing (1) and (2). Of course, computations of the box
products require determinants anyway and so there is not much saving
of time. Moreover, the identity (6) itself is not as standard as some
other identities about vectors.
The computations in this problem are simple but highly repe-
titious and prone to errors. The conceptual part is very elementary
and straightforward. All it involves is the computations of the dot
and the cross products and the norm of a vector. All these three were
also needed in the solution to Q. 12 in Section A. It is not clear what
purpose is served by this duplication.
The answer, perhaps, lies once again in the helplessness of the paper-
setters imposed by the format of the examination. In a conventional
examination, candidates can be asked to prove identities like (6). This
is precluded in a multiple choice test. Numerical problems, on the other
hand, have clear cut, crisp answers and are more suitable to be asked
in the multiple choice format. But that puts some constraint on the
topics that can be covered. So problems involving the same formulas
about the dot and the cross products of vectors appear twice.
But with a little imaginativeness, the situation could have been
salvaged. For example, instead of asking a proof of (6), the paper-
setters could have given an equation of the form
(~a × ~b) × (~c × d)
~ = α~a + β~b (7)

where α and β are some scalars. The candidates could then have been
asked to identify them from a set of four choices. Such a question could
have replaced one of the two questions based on the same formulas.

There are four comprehensions, each with
three questions with single correct answers.

Comprehension I

There are n urns each containing n+1 balls such that the i-th urn contains
i white balls and n + 1 − i red balls. Let Ui be the event of selecting the i-th
urn i = 1, 2, . . . , n, and W denote the event of getting a white ball when a
ball is drawn at random from the selected urn.

Q. 21 If P (Ui ) ∝ i for i = 1, 2, . . . , n, then lim P (W ) equals

(A) 1 (B)
3 1
(C) (D)
4 4

Answer and Comments: (B). We are given that P (Ui ) = λi for some
constant λ (independent of i). Obviously, the events U1 , U2 , . . . , Un are
mutually exclusive and exhaustive. So, their probabilities add up to 1.
This gives
n n n
X X X λn(n + 1)
P (Ui ) = λi = λ i= =1 (1)
i=1 i=1 i=1 2

from which we get λ = and hence
n(n + 1)
P (Ui ) = for i = 1, 2, . . . , n (2)
n(n + 1)

(Note that λ depends on n. Therefore it would perhaps be better to

denote it by λn . But that does not matter because now that we have
found P (Ui ) explicitly, we shall not not need λ anymore.)

The next task is to determine P (W ) as a function of n. The event W
is the disjunction of the mutually exclusive events U1 W, U2 W, . . . , Un W .
P (W ) = P (Ui W ) (3)

Now, the event Ui W is the conjunction of the events Ui and W . There-

fore by the law of conditional property, we have

P (Ui W ) = P (Ui )P (W |Ui ) for i = 1, 2, . . . , n (4)

where P (W |Ui ) is the conditional probability of W given Ui . This is

precisely the probability of drawing a white ball from the i-th urn. As
the i-th urn contains n + 1 balls of which i are white, we have
P (W |Ui ) = for i = 1, 2, . . . , n (5)
Putting (2) and (5) into (4), we get

P (Ui W ) = for i = 1, 2, . . . , n (6)
n(n + 1)2
and hence from (3),
X 2i2
P (W ) = 2
i=1 n(n + 1)
n(n + 1)2 i=1
2n(n + 1)(2n + 1)
6n(n + 1)2
2n + 1
= (7)
3n + 3
It is now immediate that lim P (W ) = . (A very special case of Ex-
n→∞ 3
ercise (6.41), part (ii).)
The problem is a hotch-potch of several parts jumbled together.
The essential idea is to express P (W ) as a function of n. (It would

be more logical to denote the event W by Wn , because it is a different
event for each n. But this might cause even more confusion.) The event
W can occur in any one of n ways, depending upon from which urn
the ball is drawn. Finding the probability of each of these n sub-events
requires conditional probability of W given Ui . Finally, the probability
of Ui is given in a rather clumsy manner. Apparently, the idea was to
test the knowledge of the formula for the sum of the first n positive

Q. 22 If P (Ui ) = c where c is a constant, then P (Un /W ) equals

2 1
(A) (B)
n+1 n+1
n 1
(C) (D)
n+1 2

Answer and Comments: (A). This problem is a little more straight-

forward at the start than the last one because P (Ui ) is given in a less
clumsy way. (Of course, instead of saying that P (Ui ) is a constant, a
better wording would have been to say that all urns are equally likely
to be selected.) Since P (Ui ) = c for every i = 1, 2, . . . , n, instead of (1)
we have a much simpler equation, viz. nc = 1 from which we get
P (Ui ) = c = for i = 1, 2, . . . , n (8)
instead of (2). Equation (3) still remains valid. Our interest now is in
determining the conditional probability P (Un /W ), i.e. the probability
that we had picked the n-th urn, it being given that the ball we picked
happens to be white. Since all urns are equally likely, it is tempting to
think that the answer is . This temptation betrays a common confu-
sion about the concept of conditional probability. (For elaboration, see
Comment No. 5 of Chapter 22.) Had been one of the given options,
chances are that some candidates would have fallen for it. Apparently,
the examiners have tried to help such candidates by alerting them that
their way of thinking is not correct.
The correct way to find P (Un /W ) is to use the law of conditional

probability. It gives
P (Un W )
P (Un /W ) = (9)
P (W )
To find the numerator, we note that the Equation (4) above still holds
because it is independent of the probabilities with which the urns are
chosen. So, applying (4) with i = n we get

P (Un W ) = P (Un )P (W/Un ) (10)

By (8) we already know P (Un ) as . As for the second factor, we note
again that Equation (5) is also still valid. So
1 n 1
P (Un W ) = × = (11)
n n+1 n+1
Let us now worry about the denominator of (9), viz. P (W ). As Equa-
tion (3) also remains valid, we have to compute P (Ui W ) for every
i = 1, 2, . . . , n. This is a generalisation of (10) with n replaced by i.
Doing the same in (11), we get
1 i i
P (Ui W ) = × = (12)
n n+1 n(n + 1)
Combining (3) and (12), we now have
n n
X X i n(n + 1) 1
P (W ) = P (Ui W ) = = = (13)
i=1 i=1 n(n + 1) 2n(n + 1) 2

So, finally, putting (11) and (13) into (9) we get P (Un /W ) = .
There is a more elegant way to arrive at the answer. Since all the
urns are equally likely to be picked, we might as well empty them all
into a single big urn and draw one ball from it at random. For each ball
in this new urn, we keep track of which urn it came from. Now this new
urn has n(n + 1) balls, of which half are white. Among these n(n + 1)
white balls, exactly n come from the n-th urn. So the probability of
n 2
Un given W is the ratio 1 which is simply .
n(n + 1) n+1

Q. 23 If n is even, P (Ui ) =for every i and E denotes the event of choosing
an even numbered urn, then P (W/E) equals
n+2 n+2
(A) (B)
2n + 1 2(n + 1)

n 1
(C) (D)
n+1 n+1

Answer and Comments: (B). Yet another computation of condi-

tional probability. But this time the problem can be taken as a con-
tinuation of the last one because the values of P (Ui ) are the same as
in the last problem. In fact, now that we know an elegant way to
solve the last problem, let us do the present one by the elegant method
only. (Those who cannot think of it are welcome to imitate the pedes-
trian solution to the last problem. In fact, this is recommended as a
drill.) So let us suppose that all the urns are emptied in a big urn
and a ball is drawn at random from it. We are given that it originally
came from an even numbered urn. Since n is even, this means that
the ball came from one of the urns numbered 2, 4, . . . , n − 2, n. And
we have to find the probability of its being white. The total number of
balls in these even numbered urns is (n + 1). The number of white
balls among them is 2 + 4 + . . . + n. Calling n/2 as m this number is
n(n + 2)
2(1 + 2 + . . . + m) = m(m + 1) = . So the desired probability
n(n + 2)/4 n+2
is simply the ratio which equals .
n(n + 1)/2 2(n + 1)

It is not clear what the word ‘comprehension’ in the title means. In the
context of examinations (especially in languages) it means that an excerpt
from some article (or a poem) is given and questions are asked to test if
the candidates have understood the passage correctly. But here, instead of
a passage, all we see is a bunch of three problems with a common setting.
Naturally, there is considerable duplication of ideas in their solutions. In fact,
Q. 22 is completely subsumed by Q. 23 and it is not clear what is gained
by this duplication. As for Q. 21, it is qualitatively different from the other
two in the bunch. But the solution does require conditional probability.
Moreover, the probabilities of the events Ui are given in an unnecessarily

clumsy manner. Perhaps, by ‘comprehension’ the paper-setters mean the
ability to make sense out of a clumsily worded problem!

Comprehension II
Rb b−a
Now we define the definite integral using the formula f (x)dx = (f (a)+
a 2
f (b)). For more accurate result, for c ∈ (a, b), we let F (c) = (f (a) +
b−c a + b Rb b−a
f (c)) + (f (b) + f (c)). When c = , f (x)dx = (f (a) + f (b) +
2 2 a 4
2f (c)).
Q. 24 sin xdx equals
π √ π √
(A) √ (1 + 2) (B) (1 + 2)
8 2 √ 8
π π √
(C) (1 + 2) (D) (1 + 2 2)
4 4

Answer and Comments: (B). The first task is to correctly under-

stand the problem. The integral that is asked here is not the usual
definite integral, which can be evaluated by first finding an antideriva-
tive, viz. − cos x of the integrand sin x. We now have to follow the
definition of the integral as given in the statement of the problem. The
trouble, however, is that the problem itself gives two different defini-
tions for f (x)dx, one at the beginning and then one at the end (for
more accurate result!) and it is not clear which of these two is to be
Rb b−a
If we follow the first definition of f (x)dx, viz. (f (a) + f (b)),
a 2
π/2 π π
then sin xdx equals (sin 0 + sin π2 ) = . But this is not given
0 4 4
as one of the alternatives. So, we interpret the question to mean the
Rb b−a
second definition, viz. f (x)dx = (f (a) + f (b) + 2f (c)). With
a 4
π/2 π/2 π 2
this definition, sin xdx = (sin 0 + sin π2 + 2 sin π4 ) = (1 + √ ) =
0 4 8 2
π √
(1 + 2) which matches with (B).
Q. 25 If f ′′ (x) < 0 for every x ∈ (a, b) and c is a point such that a < c < b,
and (c, f (c)) is the point lying on the curve for which F (c) is maximum,
then f ′ (c) is equal to
f (b) − f (a) 2(f (b) − f (a))
(A) (B)
b−a b−a

2f (b) − f (a)
(C) (D) 0
2b − c

Answer and Comments: (A). Like the last question, the present one
is straightforward once you know what it is asking. We do not know
the function f (x). But we have a new function F (c) defined in terms
of f (c) by

c−a b−c
F (c) = (f (a) + f (c)) + (f (b) + f (c)) (1)
2 2
In this expression a, b, f (a), f (b) are constants. But c varies over the
interval [a, b]. So, this expression is a function of c and we have to
maximise it for c ∈ (a, b). As f (c) is given to be differentiable as a
function of c, we can differentiate (2) w.r.t. c and get

f (a) + f (c) (c − a)f ′ (c) f (b) + f (c) (b − c)f ′ (c)

F ′ (c) = + − +
2 2 2 2
f (a) − f (b) (b − a)f ′ (c)
= + (2)
2 2
f (b) − f (a)
It follows that F ′ (c) = 0 only when f ′ (c) = . So, if at all one
of the given alternatives is correct, it must be (A). Note, incidentally,
that such a point c ∈ (a, b) does exist by the Lagrange’s Mean Value
Theorem. Moreover, it is unique because the hypothesis f ′′ (x) < 0
implies that f ′ (x) is strictly monotonically decreasing and hence a one-
to-one function. Of course, we have not proved that at F will have a
maximum at this point c. In a time constrained examination of the
multiple choice type, it would be foolish to spend time on it. But an
honest answer must verify it. To do so, we note that F has no other
interior critical point and so the second derivative test can be applied.
Therefore we shall be through if we can show that F ′′ (c) < 0 if c is such

f (b) − f (a)
that f ′ (c) = . This is easy, because differentiating (2) we
b − a ′′
have F ′′ (c) = f (c) which is always negative since we are given
that f (x) < 0 for all x ∈ (a, b).
′′ af (x)dx − t−a
(f (t) + f (a))
Q. 26 If f (x) is continuous everywhere and lim =0
t→a (t − a)3
for all a, then f (x) is a polynomial in x whose maximum possible degree
(A) 1 (B) 2
(C) 3 (D) 4

Answer and Comments: (A). Once again the correct interpretation

of the problem is half the solution. If we interpret the integral in the
numerator according to the first definition given in the statement of the
problem, then the numerator is identically zero and so the hypothesis
will always hold no matter what the function f (x) is. In this case,
no conclusion can be drawn about the degree of f (x). So we have to
discard this interpretation.
Let us, therefore, take the second definition of the integral. Then
the numerator in the problem equals
(t − a) a+t t−a
(f (a) + f (t) + 2f ( )) − (f (a) + f (t))
4 2 2
(t − a) a+t
which reduces to (2f ( ) − f (a) − f (t)) after simplification.
4 2
So our hypothesis now becomes
2f ( a+t
) − f (a) − f (t)
lim =0 (3)
t→a (t − a)2
for all a. As f (x) is given to be twice differentiable, this limit, say L,
can be evaluated applying L’Hôpital’s rule twice, giving
f ′ ( a+t
) − f ′ (t)
L = lim
t→a 2(t − a)
1 ′′ a+t
f ( 2 ) − f ′′ (t)
= lim 2
t→a 2
1 ′′
= − f (a) (4)
where in the last step we have used the continuity of f ′′ at a.
So, the hypothesis now means that f ′′ (a) = 0 for all a or, in other
words, that f ′′ is identically zero. Integrating, we get

f ′ (x) = A (5)

for some constant A. (Here, by integration we mean the ordinary inte-

gration and not the one given in the problem!) One more integration

f (x) = Ax + B (6)

for some constants A and B. Hence f (x) is a polynomial of degree 1

or 0 (depending upon whether A 6= 0 or A = 0).

Because of the unavailability of the original question paper, it is not clear

how the preamble to this section was worded. Instead of defining f (x)dx
as (f (a) + f (b)), a better wording would have been to say that the
latter expression is an estimate of the ordinary definite integral f (x)dx.
Also, some motivation of how F (c) is arrived at could have been given. It
is obtained by splitting the original integral f (x)dx into two parts, viz. as
Rc Rb
f (x)dx+ f (x)dx and then applying the earlier estimate to each of the two
a c
integrals. In general, F (c) would be a more accurate estimate of the integral
f (x)dx if the intermediate point c is suitably chosen. But the right choice
will depend on the function f and the interval [a, b]. In absence of any other
information about f , one standard choice for c is to take it as the mid-point
of the interval [a, b].
Had the preamble explained these things, the three questions would have
been less obscure. The first question (Q. 24) simply asks the (revised) es-
timate for a particular definite integral. The hypothesis in Q. 25 makes
the function strictly concave downwards on the interval [a, b] (see Comment
No. 18 of Chapter 13). In this case, no matter how you choose c ∈ (a, b),
the estimate F (c) will always fall short of the exact value of the integral

f (x)dx. The significance of the problem is that it tells you that the
best possible estimate is obtained by taking c to be the ‘mean value’ of
f (x) over [a, b] as given by the Lagrange Mean Value Theorem. The third
question (Q. 26) deals with the difference between the revised, more accu-
rate estimate (f (a) + f (b) + 2f ( a+b
)) and the original crude estimate
(f (a) + f (b)). As the length of the interval tends to 0, this difference
obviously tend to 0. The hypothesis deals with how rapidly it tends to 0.
(See Comment No. 6 of Chapter 15 for an elaboration of this concept.)
The problem amounts to showing that unless the function f (x) is a linear
function, the difference between these two estimates cannot tend to 0 more
rapidly than the cube of the length of the interval of integration.
The topic in this section
is very interesting. The standard method to eval-
uate a definite integral a f (x)dx is to first find an antiderivative, say F (x),
of the integrand and then evaluate the integral as F (b) − F (a). However,
as explained in Comment No. 25 of Chapter 17, many times we run into
situations where we cannot express the antiderivative in a closed form, as
happens, for example, when f (x) = ex or sin(x2 ). In such cases we have
to resort to methods such as the trapezoidal rule and the Simpson’s rule to
evaluate the integral approximately. The two estimates given in the present
section are among the simplest such approximations.
In fact, instead of giving some numerical problems, the paper-setters could
have asked some questions to test the ability of the candidate to digest con-
cepts. Here are some examples, along with the expected answers:

1. What is the geometric significance of the first estimate of the

integral, viz. (f (a) + f (b)) in relation to the graph of f ?
Ans: It is the area of the trapezium with vertices (a, 0), (b, 0), (a, f (a))
and (b, f (b)). The last two vertices lie on the graph of the function f
while the first two are their projections on the x-axis. So, the given
estimate is the area below the chord of the graph of the function joining
the points (a, f (a)) and (b, f (b)) on the graph. It is approximately the
area below the portion of the graph of f (x) between x = a and x = b.

2. How is the revised estimate F (c) related to the first estimate?

Ans: It is obtained by splitting the original integral f (x)dx into two
Rc Rb
parts, viz. as f (x)dx+ f (x)dx and then applying the earlier estimate
a c
to each of the two integrals.
3. How is the first estimate related to the ordinary definite inte-
gral f (x)dx defined in terms of the limit of Riemann sums?
Ans: Both f (a)(b−a) and f (b)(b−a) are certain Riemann sums of the
function f (x) for the partition of [a, b] into a single interval, viz. [a, b]
itself. The first estimate is the arithmetic mean of these two Riemann
4. If f (x) is concave upwards on [a, b], how does the revised esti-
mate F (c) compare with the exact value of the integral f (x)dx?
Also, how does F (c) compare with the first estimate?
Ans: Since f (x) is concave upwards, the chords always lie above the
graphs. Therefore, no matter which c ∈ (a, b) is chosen, the estimate
F (c) always exceeds the integral, i.e. f (x)dx < F (c). But we also
have F (c) < (f (a) + f (b)). So, it is a better estimate of the
integral than the first estimate.
5. For which functions f (x) will the estimate (f (a) + f (b)) be
exact (i.e. coincide with the integral f (x)dx) for every choice
of the interval [a, b]?
Ans: When f (x) is a linear function, i.e. a function of the form f (x) =
Ax + B where A, B are some constants.
Had some such questions been designed, they would have tested ‘compre-
hension’ in the true sense of the term. It is a pity that the ‘comprehension’
in this bunch as it now stands amounts to trying to make some sense out of
some obscure, clumsily stated problems.

Comprehension III

Let ABCD be a square of side length 2 units. C2 is the circle through
vertices A, B, C, D and C1 is the circle touching all the sides of the square
ABCD. L is a line through A.

P A2 + P B 2 + P C 2 + P D 2
Q. 27 If P is a point on C1 and Q is a point on C2 , then
QA2 + QB 2 + QC 2 + QD 2
(A) 0.75 (B) 1.25
(C) 1 (D) 0.5

Answer and Comments: (A). An excellent example of how a mul-

tiple choice question should not be. The real crux of the problem is
to show that the ratio in the statement of the problem is constant, i.e.
independent of the points P and Q chosen, as long as they lie on the
circles C1 , C2 respectively. Once this is shown, finding the numerical
value of this ratio is a relatively minor matter. In fact, it will come
built-in in the proof of constancy of the ratio.
But since it is a multiple choice question, a smart student may
simply assume that the ratio is in- y

dependent of the particular C 2

points P and Q. To find its B 1 1 A=Q

numerical value, he can choose 1

C 1
P and Q to be any points for
P x
which both the numerator and O
the denominator can be calcu-
lated easily. One such choice is C 1 1 D

to take P as a point of contact

of C1 with a side of the square
as shown in the figure and take
Q as the vertex A itself.

With this choice we have P A = P D = 1 and P B = P C = 5.

P A2 + P B 2 + P C 2 + P D 2 = 1 + 5 + 5 + 1 = 12 sq. units (1)

√ √
Similarly, QA = 0, QB = QD = 2 and QC = 4 + 4 = 8. Hence

QA2 + QB 2 + QC 2 + QD 2 = 0 + 4 + 4 + 8 = 16 sq. units (2)

Therefore, the ratio asked in the problem is 16
= 0.75. Hands down!
Fortunately, the work involved in an honest solution to the
problem is not prohibitively lengthy either. With a suitable choice
of coordinates as shown in the figure, take the vertices A, B, C, D as
(1, 1), (−1, 1), (−1, −1) and (1, −1) respectively. Let P = (x1 , y1 ) be
any point on the circle C1 and Q = (x2 , y2) be any point on the circle
C2 . Then we have

x21 + y12 = 1 (3)

and x22 + y22 = 2 (4)

Now a straightforward computation gives

P A2 = (x1 − 1)2 + (y1 − 1)2 = 3 − 2x1 − 2y1 (5)

P B 2 = (x1 + 1)2 + (y1 − 1)2 = 3 + 2x1 − 2y1 (6)
P C 2 = (x1 + 1)2 + (y1 + 1)2 = 3 + 2x1 + 2y1 (7)
and P D 2 = (x1 − 1)2 + (y1 + 1)2 = 3 − 2x1 + 2y1 (8)

where we have used (3)

Adding these four equations, we have

P A2 + P B 2 + P C 2 + P D 2 = 12 (9)

By an entirely analogous computation using (4), we get

QA2 + QB 2 + QC 2 + QD 2 = 16 (10)

(9) and (10) together give the answer. Actually, we have shown a little
more than the problem asks. We have not only shown that the ratio in
the problem is a constant, but we have shown that the numerator as
well as the denominator are constants (i.e. independent of P and Q).
It is obvious that this is the way the paper-setters intended the
problem to be solved. Otherwise there was no need to give the side
of the square as 2 units length. Indeed the answer is independent of
the side of the square. Had the side been some other length, then
by a similarity transformation all lengths would get multiplied by the
same constant of proportionality. Therefore their squares would get

multiplied by the square of this constant. As a result the ratio in the
problem would remain intact.
By giving the side of the square as 2 units, the paper-setters have
apparently tried to help the honest students. This is commendable.
But what they probably failed to realise is that a crook does not need
their help and can get the answer in the sneaky manner given at the
beginning. Because the honest solution is also easy, the resulting saving
may be only a minute or so. But in a severely time-constrained, highly
competitive examination, one minute is significant.
But then again, perhaps the paper-setters did think of the possibility
of sneaking but decided to let it pass. A candidate who can think of
it is probably intelligent enough to do the problem in the conventional
way too. In fact, from a practical point of view he is more intelligent.
And it is possible that the paper-setters wanted to reward this quality.
Although nobody is likely to do this problem by pure geometry,
we give here one such solution for the lovers of pure geometry (a fast
dying creed). We shall prove separately that both the numerator and
P A2 + P B 2 + P C 2 + P D 2
the denominator of the ratio are constants
QA2 + QB 2 + QC 2 + QD 2
(i.e. independent of P and Q, as long as P lies on C1 and Q on C2 ).
Interestingly, it is easier to do this for the denominator because then Q
lies on the circumcircle of the square ABCD. Since AC and BD are
diameters of this circle, the triangles AQC and BQD are both right-
angled. (We do not draw a separate diagram here. We refer the reader
to the diagram in the earlier solution and ask him to do the necessary
fillings.) So, by Pythagoras theorem,

QA2 + QC 2 = AC 2 = 4r22 (11)

and QB 2 + QD 2 = BD 2 = 4r22 (12)

where r2 is the radius of C2 . Adding these two equations we get that

the denominator is a constant equal to 8r22 (which equals 16 in the
present problem).
But this argument cannot be applied for the numerator because
the 6 AP C and 6 BP D need not be right angles. In this case we need
an extension of the Pythagoras theorem called Apollonius theorem (see
the end of Comment No. 4 of Chapter 11). The point O is the midpoint

of both AC and BD. So P O is a median of both ∆AP C and ∆BP D.
So, Apollonuis theorem gives the following analogues of (7) and (8)
P A2 + P C 2 = AC 2 + 2OP 2 = 2r22 + 2OP 2 (13)
and P B 2 + P D 2 = BD 2 + 2OP 2 = 2r22 + 2OP 2 (14)
Adding these two equations we see that the numerator equals 4r22 +
4OP 2 (which equals 8 + 4 i.e. 12 in the present problem.)
Actually, the pure geometry solution is not substantially different
from the earlier one based on coordinate geometry. In the earlier solu-
tion, we added the four Equations (5) to (8) together. Instead, suppose
we first add only (5) and (7). Then we get

P A2 + P C 2 = 3 (15)

which is a special case of (13) with r2 = 2 and OP = 1. Similarly,
(14) is the equivalent of adding (6) and (8). So, in principle, the two
solutions are not different. In the coordinate geometry proof, when we
add P A2 and P C 2 , certain terms cancel because of algebra. In the
pure geometry solution, we need the Apollonius theorem to add them.
But if you look at the proof of the Apollonius theorem, it is essentially
algebraic, because it consists of expressing each of P A2 and P C 2 using
the Pythagoras theorem and then doing some cancellation.

Q. 28 A variable circle touches the line L and the circle C1 externally and
both the circles are on the same side of the line L. Then the locus of
the centre of the circle is
(A) an ellipse (B) a hyperbola
(C) a parabola (D) parts of a straight line

Answer and Comments: (C). In the preamble of this comprehen-

sion, L is given to be a line through the vertex A of the square ABCD.
This line figured nowhere in the last problem. In this problem, there is
a reference to it with a further restriction that the circle C1 lies entirely
on one side of L. This can be confusing to a candidate. If after reading
the preamble, he has made a rough sketch in which the line L cuts the

circle C1 (which was perfectly legitimate to do as nothing was specified
about the line L other than that it passed through A), he will now have
to draw another sketch in which L is completely outside C1 , thereby
resulting in an unnecessary waste of precious time. It would have been
far better to say nothing about L in the preamble and introduce it only
now along with the restriction on it.
As for the circle C1 , in the preamble it was given to be the incircle of
a certain square of side 2 units length. But that introduction is totally
irrelevant in the present problem. Basically, the present problem is
about any fixed circle C1 and any line L not cutting it. The problem
deals with the locus of the centre of a variable circle which lies on the
same side of L as C1 does and touches the line L and also the circle C1
externally. Note that since the equations of C1 and L are not given,
we cannot determine the locus completely. But we can determine what
type of a curve it represents and that is all that the problem asks.
Let M and r1 be the centre and the radius of the fixed circle and
P and r be the centre and the radius of the variable circle C. We want
to find the locus of P . Because y

of the data, MP = r1 + r.
Also, the perpendicular dis- C1

tance of P from L is r. If M (0, b)

the two distances were equal C
P(h , k)
then the locus of P would
have been a parabola with fo- L
O x

cus M and directrix L. But r1

as it stands the distance of P L’

from the line L falls short of

its distance from the point M
by a fixed constant, viz. r1 .
This difficulty can be resolved neatly by finding another fixed

line L whose distance from P is r + r1 . The correct choice is to let
L′ be the line parallel to L at a perpendicular distance r1 from it and
lying on the opposite side of L as the circle C1 . We now see without
any computations that the locus of P is a parabola with focus M and
directrix L′ .
Note that there is hardly any computation in the solution. The
success depends on the construction of the line L′ . That puts this

particular problem in a class above the rest. Conceiving such elegant
constructions is akin to an artistic experience and the ability to come
up with them is a valuable asset. In pure geometry problems, there is
ample scope for this ability.
What if you cannot think of this trick on the spur of the moment?
Then everything is not lost. The ever so reliable coordinates are always
available for help. Take the given line L as the x-axis and choose the
y-axis to pass through the point M, the centre of the given circle C1 .
Then we can take M as (0, b) where we may suppose b > 0 without
loss of generality. As is a usual practice in locus problems, let (h, k) be
the current coordinates of the point P . As P and M lie on same side
of L, we have k > 0. Therefore, the distance of P from L is k and this
is also the radius of the circle C. As C touches C1 externally, we have
h2 + (k − b)2 = r1 + k (16)

On squaring and simplifying, this becomes h2 −(2b+2r1 )k +b2 −r12 = 0.

Hence the locus of P is x2 −(2b+2r1 )y +b2 −r12 = 0 which is a parabola
as the quadratic terms form a perfect square. Actually, one need not
even do this much. The question does not ask you to identify the locus
but the type of curve it represents. It is clear just by inspecting (16)
that when you square it the k 2 terms on both the sides will cancel
out. As the only other quadratic term is h2 , the curve is going to be a
Of course, this solution is not as elegant as the earlier one. But that’s
the very price you pay for the reliability of coordinates. In the case of
the last problem, we remarked that there was no substantial difference
in the pure geometry proof and the one based on coordinates. But in
the present problem, the coordinate geometry solution is no match to
the pure geometry solution in terms of elegance.
The problem is of the same spirit as a 2001 JEE question (see
Exercise (9.15 (c)) where you had to identify the locus of the centre
of a variable circle, which touches two given circles, one internally and
the other externally. But its solution was not so tricky. That ques-
tion carried 5 points in a two hour paper with total 100 points. On a
proportionate basis that gives 6 minutes. The present problem is for 5
points in a two hour paper with 184 total marks. So, now you get con-

siderably less time. Of course, you do not have to write your reasoning
Q. 29 A line M through A is drawn parallel to BD. A point S moves such
that its distances from the line BD and the vertex A are equal. If the
locus of S cuts M at T2 and T3 and AC at T1 , then the area of ∆T1 T2 T3
1 2
(A) sq. units (B) sq. units
2 3
(C) 1 sq. units (D) 2 sq. units

Answer and Comments: (C). Another problem on parabola. The

locus of S is a parabola with focus A and directrix the line BD. As
AC is perpendicular to BD and hence to the directrix M, it follows
that the axis of the parabola is along AC. Let O be the centre of this
square. Then T1 is the midpoint of
OA. (This fact was also 3

used in the solution to Q. M

10.) The points T2 and B A

T3 lie on the parabola and

also on the line M. As the T

perpendicular distance be- O


tween M and BD is AC,

the perpendiculars from T2 C D

and T3 to BD must meet

at D and B (or vice versa).
So ∆T1 T2 T3 is an isosceles
triangle with base T2 T3 . So
its area is simply AT1 .AT2 .
It only remains to find these lengths. From the preamble, √ the
side of the square√is given as 2 units length. So its diagonal is 2 2.
Therefore OA = 2. Since AT1 = 21 OA while AT1 = OA, we get the
area of the triangle T1 T2 T3 as 12 OA2 = 1 square units. The segment
T2 T3 is popularly called the latus rectum of the parabola and we could
have used the fact that its length is four times the distance between the
vertex T1 and the focus A. But that does not simplify the calculations.
Unlike in the last comprehension, the three problems in the present one
are totally unrelated to one another. The last problem has considerable du-

plication of ideas with Q. 10. As there is no common theme to the three
problems in the present bunch, it is difficult to comprehend what ‘compre-
hension’ means. In fact, as pointed out at the beginning of the solution to Q.
28, introducing the line L in the preamble and then imposing a restriction
on it later can only result in a waste of time.

Comprehension IV
 
1 0 0
Let A =  2 1 0 

 and U1 , U2 , U3 be column matrices satisfying the
3 2 1
     
1 2 2
equations AU1 =  0  , AU2 =  3  and AU3 =  3 . Let U be the 3 × 3
     

0 0 1
matrix whose columns are U1 , U2 , U3 .

Q. 30 The value of |U| is

(A) 3 (B) −3
(C) 3/2 (D) 2

Answer and Comments: (A). The straightforward way is to begin

by identifying each of the column vectors U1 , U2 , U3 . Each requires us
to solve a system
 of three
 equations in three unknowns.
  For example,
u11 1
suppose U1 =  u21 . Then the equation AU1 =  0  is equivalent
   

u31 0
to the system

u11 = 1 (1)
2u11 + u21 = 0 (2)
3u11 + 2u21 + u31 = 0 (3)

which can be solved by inspection to get

 
U1 =  −2  (4)
 

Similar calculations yield
   
2 2
U2 =  −1  and U3 =  −1 
  
 (5)
−4 −3

Put together this gives the matrix U as

 
1 2 2
U =  −2 −1 −1 

 (6)
1 −4 −3

Now that we have identified the matrix U explicitly, we can answer any
questions about it. The present problem asks for its determinant |U|.
A straightforward calculation gives

|U| = 1 × (−1) + 2 × (−7) + 2 × 9 = 3 (7)

A slightly better approach is to put together the three equations given

in the statement of the problem to give the single matrix equation

AU = B (8)

where A and B are the matrices given by

 
1 0 0
A= 2 1 0 

 (9)
3 2 1

 
1 2 2
B= 0 3 3 

 (10)
0 0 1

The matrix A is non-singular since its determinant is 1 which is non-

zero. Therefore A−1 exists and so from (8) we get

U = A−1 B (11)

This gives us an alternate method for calculating the matrix U. But

the computations involved in finding A−1 are not substantially different

from those in solving the three equations (1), (2) (3) simultaneously.
Moreover, we still have to multiply the two matrices A−1 and B. So
there is not much saving in this approach.
But there is a slicker way to find |U| even without finding U.
Taking determinants of both the sides in (8),

|A||U| = |B| (12)

From (9) we know |A| = 1. Also, from (10) we compute |B| = 3.

Putting this into (12) gives |U| = 3. This solution is certainly more
elegant in terms of its approach, which is based on the multiplicativity
property of determinants (i.e. the fact that the determinant of a prod-
uct of two matrices is the product of the determinants.) Surely, it is
much easier to multiply determinants (which are some real numbers)
than to multiply two 3 × 3 matrices which is a nasty job.
Of course, evaluating a 3 × 3 determinant simply by expanding
all terms is not a very exciting job either. But in this particular prob-
lem, this part too is drastically simplified because of certain particular
features of the matrices A and B. Note that for the matrix B all the
entries below the main diagonal are zero. A matrix with this property
is called upper triangular. Similarly, the matrix A given by (9) is
lower triangular because all entries above the main diagonal vanish.
In either case, the determinant of the matrix has only one non-zero
term, viz. the product of the diagonal elements. So the determinants
of B and of A can be written down simply by inspection.
This problem is excellent because it rewards those who look for
elegant solutions before proceeding with brute force computations.

Q. 31 The sum of the elements of U −1 is

(A) −1 (B) 0
(C) 1 (D) 3

Answer and Comments: (B). Once again, the most straightforward

approach would be to begin by calculating the matrix U −1 explicitly.
If we have already identified U by (6) in the solution to the previ-
ous problem, then using the standard formula for the inverse in terms
of the cofactors of the elements of a matrix, one can show, by sheer

computation, that
 
−1 −2 0
adj (U) 1
U −1 = =  −7 −5 −3  (13)

|U| 3
9 6 3

from which it is trivial to verify that the sum of the elements of U −1 is

But the computations involved in finding U −1 , while inherently
simple, are highly repetitious and prone to errors. So once again we
look for some better method. Note that we are not asked to find the
matrix U −1 per se but only the sum of its elements. Just as in the last
problem, we were able to find the determinant of the matrix U even
without finding U explicitly, let us see if we can do something similar
We begin by an observation which is applicable to any square
matrix C = (cij ) of order 3 (or of any order n for that matter). It is
  thatthe columns
  of C are precisely the product matrices
1 0 0
C  0 , C  1  and C  0  respectively. Adding these three, we get
     

0 0 1
   
1 c11 + c12 + c13
C   =  c21 + c22 + c23 
1 (14)
   

1 c31 + c32 + c33

 
Verbally, multiplying a matrix on the right by the column vector  1 
 

gives a column vector whose entries are the row sums of the matrix C.
An entirely analogous argument shows that [1 1 1]C is a row vector
whose entries are the column sums of the matrix C.
What if we perform both the operations, i.e. premultiplication
by the row vector [1 1 1] and also post-multiplication by the column
vector  1 ? Then, as expected, we get a 1 × 1 matrix whose lone
 

entry is the sum of all the entries of C. It is customary to denote a
1 × 1 matrix by its lone entry. We then have
 
1 X3 X 3

[1 1 1] C   =

cij (15)
1 i=1 j=1

This simple formula allows us to recast our goal. We want the

sum, say S, of the elements of the matrix U −1 . Because of (15) we can
rewrite this as
 
−1 
S = [1 1 1] U  1  (16)

Of course, merely recasting a problem does not solve it! Indeed,

if we have already calculated the matrix U −1 , then it is foolish to use
(16). Instead, it is far better just to look at the entries of U −1 and add.
The true worth of a formula like (16) comes when we have not explicitly
calculated the matrix U −1 but can nevertheless identify it in some other
way, e.g. as the product of some simpler matrices. This is indeed the
case in the present problem. For, since both A and B are invertible,
from Equation (8) we have U = A−1 B and hence U −1 = B −1 A. Putting
this into (16), we now have
 
−1 
S = [1 1 1] B A  1 
 (17)
Here S, which is simply a real number, is expressed as a product of
four matrices! Using associativity of matrix multiplication we are free
to group any of these together as long as we do not change their order.
In particular, we can rewrite S further as
S = PQ (18)
P = [1 1 1] B −1 (19)
 
and Q = A  1  (20)
 

We already know that Q is a column vector whose entries are the row
sums of A. From (9), we have
 
Q= 3  (21)
 

Unfortunately, P cannot be computed so readily. We know it is a

row vector whose entries are the column sums of the matrix B −1 . The
trouble is that even though we know B from (10), we have not yet found
its inverse. It may thus appear that we have not really gained anything
by this approach over the earlier approach. The only difference is that
instead of the matrix U we now have to find the inverse of the matrix B.
But this is no small gain. First, we are given B whereas we have to find
U (which so far we have escaped). Secondly, even if we compute U as in
(6), computing its inverse is considerably more laborious than finding
the inverse of B. This happens again because B is an upper triangular
matrix. It is easy to show that the inverse of every such matrix (if it
exists) is also upper triangular and further that its diagonal entries are
precisely the reciprocals of the corresponding diagonal entries of the
original matrix. Even if we do not know this result, we can find B −1
simply by solving some equations.
 
x1 y1 z1
So, let B −1 =  x2 y2 z2 . Then BB −1 = I gives
 

x3 y3 z3
   
x1 + 2x2 + 2x3 y1 + 2y2 + 2y3 z1 + 2z2 + 2z3 1 0 0

 3x2 + 3x3 3y2 + 3y3 3z2 + 3z3  =  0 1 0 
 
x3 y3 z3 0 0 1

which is a system of 9 equations in 9 unknowns. A similar system for

finding U −1 would be horrendous. But not the present one. For we
immediately get x3 = 0, y3 = 0, z3 = 1 and thereafter x2 = 0, y2 =
1/3, z2 = −1 and finally the first row entries, viz. x1 = 1, y1 = −2/3
and z1 = 0. Hence
 
1 −2/3 0
B −1 =  0 1/3 −1  (23)
 

0 0 1

(Note, incidentally, that B −1 is upper triangular as we had remarked
it would have to be and, further also, that its diagonal entries are
the reciprocals of the corresponding diagonal entries of B. In fact,
those who know this already could have incorporated this information
to reduce the number of unknowns in (22) from 9 to 3 and thereby
simplify the calculations still further.)
Putting (23) into (19), we get

P = [1 − 1/3 0] (24)

Finally, putting (18), (21) and (24) together, we have

 
S = P Q = [1 − 1/3 0]  3  = 0 (25)
 

which completes the solution.

As in the last problem, in this problem too, the gap of labour
needed between the elegant and the brute force solutions is remark-
able. But the difference is that in the last problem the key idea in
the elegant solution was a familiar one, viz. the multiplicativity prop-
erty of determinants. And once you hit it, hardly any numerical work
had to be done. Comparatively, the key idea behind the elegant solu-
tion to the present problem, viz. Equation (15), is not so frequently
used and hence not easy to come up with. Moreover, you have to do
some numerical work to get B −1 anyway. As a result, even though the
present problem is also very good, it is a shade below the last one. True
elegance lies in using something usual in an unusual way.
 
Q. 32 The value of [3 2 0] U  2  is
 

(A) 5 (B) 5/2
(C) 4 (D) 3/2

Answer and Comments: (A). In the last problem, even if we had

calculated the matrix U explicitly, there was considerable more work
to do in the brute force method to find the sum of all the entries of

U −1 . In the present problem, on the other hand, if we have calculated
U already, then the best way to get the answer is to carry out the
multiplications. Thus,
    
3 1 2 2 3
[3 2 0] U  2  = [3 2 0]  −2 −1 −1   2 
    

0 1 −4 −3 0
 
= [−1 4 4]  2 
 

= −3 + 8 = 5 (26)

In the last two problems, we could avoid the computation of the

matrix U by resorting to some elegant tricks. There seems to be no
way to do so in the present problem. The only thing we can possibly
try is to write U as A−1 B by (11). Then reasoning as in the elegant
solution to the last problem, the desired number, say T , equals
   
3 3
T = [3 2 0] U  2  = [3 2 0] A−1 B  2 
   

0 0
= LM (27)


L = [3 2 0] A−1 (28)
 
and M = B  2  (29)
 

As we already know B by (10), computation of M is immediate.

    
1 2 2 3 7
M =  0 3 3  2  =  6  (30)
    

0 0 1 0 0

The computation of L is also equally easy. But first we must find A−1 .
As A is a lower triangular matrix, its inverse can be found by a method
which is analogous to the one we used in the last problem for finding

the inverse of B which was upper triangular. In fact, we now use the
comments made there after finding B −1 to start with a considerably
simplified format for A−1 . Since we already know that A−1 is going
to be lower triangular and also that its diagonal entries will be the
reciprocals of the corresponding diagonal entries of  A, right at the
1 0 0
start, we take A−1 in a highly simplified form, viz.  a 1 0  where
 

b c 1
we have only three unknowns, viz. a, b and c. Then the requirement
AA−1 = I spells out as
   
1 0 0 1 0 0

 2+a 1 0 = 0 1 0 
 
 (31)
3 + 2a + b 2 + c 1 0 0 1

from which we get a = −2, c = −2 and b = 1. Therefore

 
1 0 0
A−1 =  −2 1 0 

 (32)
1 −2 1

We are now in a position to compute L using (28), giving

L = [3 2 0] A−1
 
1 0 0
= [3 2 0]  −2 1 0 
 

1 −2 1
= [−1 2 0] (33)

Putting (33) and (30) into (27), we finally get

 
T = LM = [−1 2 0]  6 

= −7 + 12 = 5 (34)

which is the same answer as before. Unlike in the last two problems, in
the present problem, the difference between the two approaches (one
with finding U and the other without finding it explicitly) is not very

dramatic. The fact that the row vector and the column vector appear-
ing in the statement of the problem are transposes of each other would
have been helpful had the matrix U been symmetric. In particular,
this would have been the case if A−1 equaled the transpose of B (or,
equivalently, B −1 were the transpose of A). For, in that case,
 
the ex-
pression asked would have been of the form [3 2 0] B B  2  which
 

means the row vector L above would have been the transpose of the
column vector M. But in that case LM would have been simply the
sum of the squares of the entries of M. Summing up, if B −1 (which
we calculated in (23)) had come out to be the transpose of A (which
is given to us in the preamble), then there would indeed have been a
significant short cut to the answer. (It is possible that the data in the
original problem was so designed as to make U symmetric, but some
of the figures got changed in its retrieval.)
The simplifications in the second problem in this comprehension
were due to the fact that that the matrix U could be expressed as
A−1 B where the matrices A−1 and B are, respectively, lower and upper
triangular. Such a factorisation of a square matrix is called a LU-
decomposition of it. Here L stands for ‘lower triangular’ and U for
‘upper triangular’. If the matrix is symmetric then the two factors can
further be chosen to be the transposes of each other. It takes some work
to find an LU-decomposition of a matrix. But once it is obtained, it
considerably simplifies the solution of systems of linear equations.

Of all the four comprehensions, the bunch of the problems in the last one
is the best. The three problems are intimately related to each other unlike
in the comprehension about geometry, where there is hardly any common
theme. There is a common theme in the problems about approximations of
definite integrals. But in absence of any elaboration of it in the preamble,
the problems look obscure and clumsy.
Sadly, the paper-setters have not done justice to the purpose of ‘compre-
hension’ as it is understood in the context of examinations. Traditionally,
examinations like the JEE are highly problems oriented. As a result, while
preparing for them, the students concentrate heavily on solving problems,
often by analogy with some familiar problems from the huge banks of solved

problems at their disposal. They often totally ignore the ‘theory’ or the
thought that goes into the solutions. So the adept ones among them can
pull every conceivable trick needed to solve a problem but often fail to com-
prehend even half a page of a mathematical text expounding some concepts.
Comprehensions were expected to correct this anomaly. None of the four
comprehensions given here does this job. Each is as problem-oriented as the
rest of the paper.

Answers are to be given as four-digit integers.

Q. 33 If the roots of the equation x2 − 10cx − 11d = 0 are a, b and those of

x2 − 10ax − 11b = 0 are c, d (where a, b, c, d are distinct numbers), then
the value of a + b + c + d is ........ .

Answer and Comments: 1210. The given conditions readily trans-

late into a system of four equations in the four unknowns a, b, c, d, viz.

a+b = 10c (1)

ab = −11d (2)
c+d = 10a (3)
cd = −11b (4)

So the straightforward approach to the solution would be to solve this

system for a, b, c, d and add their values. But one has to be wary that
in doing so one is not aiming too high. The focus should be on what is
asked. (See the first tip given in Comment No. 3 of Chapter 24.) Here
we are not asked the individual values of a, b, c, d but only their sum.
So there may be a simpler way to get a + b + c + d without first finding
each of a, b, c, d.
Indeed the symmetry of the equations suggests one such way. If
we add (1) and (3) we get

b + d = 9(a + c) (5)

So, we would be through if we can find a + c. Similarly, if we multiply

(2) and (4), we get abcd = 121bd. If b and d are both non-zero, this

ac = 121 (6)

What happens if either b or d vanishes? Suppose, for example that

b = 0. Then by (2), d would also vanish, contradicting that a, b, c, d

are given to be distinct. Similarly, vanishing of d would force that of b
too by (4), and hence a contradiction. So, (6) is true. (This is a sad
feature of the objective type tests. A scrupulous student will spend
time eliminating the degenerate possibilities. An easy going candidate
who cannot even think of them gets rewarded in terms of the time
saved, because no reasoning is to be shown. Only the final answer
To get more relationship between a and c, we use the fact that a
is a root of the first quadratic and c is a root of the second. Coupled
with (6), these give, respectively,

a2 = 1210 + 11d (7)

and c2 = 1210 + 11b (8)

(We could also have gotten (7) by multiplying (1) by a and then us-
ing (2) and (6). Similarly, there is an alternate derivation for (8).
Mathematically, this is not surprising because the equations (1) to (4)
together are equivalent to the data in the problem. But depending on
the inclination of a particular student, it may happen that one method
strikes him as more natural.)
We add these (7) and (8) and further add 2ac (=242) to get

(a + c)2 = 2652 + 11(b + d) = 99(a + c) + 2652 (9)

where we have used (5) too. The problem is now essentially solved.
We treat (9) as a quadratic in (a + c), solve it and then by (5) get the
value of b + d too.
But the details need to be attended to. In writing the discriminant
of the quadratic, it is foolish to expand (99)2 . Instead, let us not
forget that the figure 2652 is divisible by 121 from its very construction
and 121 is the square of 11 which is a factor of 99. Because of these
simplifications, we get
q √
99 ± 11 (9)2 + 4 × 22
99 ± 11 169
a+c = = (10)
2 2
99 ± 11 × 13 11(9 ± 13)
= =
2 2
= 121 or − 22 (11)

Once again, a scrupulous student will worry what happens if a + c =
−22. In that case, because of (6), we shall get a = c = −11. But this
contradicts that a, b, c, d are all distinct. So he will conclude that

a + c = 21 (12)

The unscrupulous one will, naturally, take the express train to (12). If
at all the thought of considering the possibility a + c = −22 occurs to
him, he will discard it on the practical ground that the answer is to be
filled in must be a positive integer!
After reaching (12), both the scrupulous and the unscrupulous
candidates will use (5) to get b + d = 121 × 9. The stupid ones will
compute this to 1089. The smart ones will retain it as it is and add
it to 121 to get 121 × (9 + 1) = 1210 as the value of a + b + c + d.
Ultimately all get the same answer!
The problem is a reasonable problem on quadratic equations
in a conventional type examination, where some partial credit could
have been reserved for dismissing the degenerate cases bd = 0 and
a + c = −22. In an examination where it is only the final, numerical
answer that matters, a sincere student who ponders on such fine points
effectively wastes his time. To discourage sloppiness, occasionally there
should be some problems where the answer lies in the degenerate cases.
5050 (1 − x50 )100 dx
Q. 34 The value of is ..... .
(1 − x50 )101 dx

Answer and Comments: 5051. Evidently, it is unthinkable to eval-

uate the integrals individually by expanding the integrands by the bi-
nomial theorem. We must resort to some trick to evaluate them. But
as in the last problem, it pays to carefully focus on what is asked. The
problem merely asks for the ratio of the two integrals rather than their
individual values. So what really matters is the mutual relationship of
the two integrals. They are strikingly similar except for the exponents
occurring in their integrands. Moreover these exponents are consecu-
tive integers.

All this suggests that each integral is a member of a sequence
of similar integrals and we are to find a formula for the relationship
between two consecutive terms of this sequence. Such a relationship is
popularly called a reduction formula. They are discussed in Chapter
This gives one possible line of attack. For every positive integer n
we let
Z 1
In = (1 − x50 )n dx (1)

The problem asks us to evaluate the ratio . This gives an implicit
hint that the reduction formula for In must be of the form where the
ratio In /In+1 is expressed as a function of n, say f (n). Our interest is in
f (100). The factor 5050 in the numerator is evidently inserted to make
the answer come out to be a whole number. It gives a clue that f (100)
should be a rational number with 5050 (or some factor of it) in its
denominator. As there is nothing special about 100, we expect, more
generally, that f (n) should come out to be a ratio of two polynomial
expressions in n. Moreover, since 5050 = 50 ×101, in general we expect
that the ratio In /In+1 should have a factor n + 1 in the denominator.
With this preamble, we now proceed to obtain a reduction formula
for In = (1−x50 )n . To relate In+1 to In , we must decrease the exponent
n+ 1 in the integrand of In+1 by 1. For this we have to take derivatives.
So, integration by parts is the right tool. If we apply it, we get
Z 1
In+1 = (1 − x50 )n+1 dx
1 Z 1
= x(1 − x50 )n+1 + 50(n + 1) x50 (1 − x50 )n dx

0 0
Z 1
= 50(n + 1) x (1 − x50 )n dx

The trouble is that the new integral we got has a factor of x50 besides
the power (1 − x50 )n in its integrand. So it is not quite In . But we can
come out of this difficulty if we rewrite x50 as (x50 − 1) + 1 and hence
rewrite the product x50 (1 − x50 )n as (1 − x50 )n − (1 − x50 )n+1 . This is

precisely the difference of the integrands in In and In+1 . As a result,
we can reduce (2) further to get

In+1 = 50(n + 1)(In − In+1 ) (3)

Or, in a simplified form,

In 50(n + 1) + 1
= (4)
In+1 50(n + 1)

Putting n = 100, = 5051 which is what is asked.
The problem is reasonable once you get the key idea needed in it,
viz. the reduction formulas, which is not difficult to arrive at. But the
paper-setters have played a little dirty game by giving the integrand
as a power of (1 − x50 ). Obviously, the exponent 50 is related to the
numbers 100 and 101 in a very obvious way. But the solution is equally
applicable if we replace 50 by any positive integer m. The only change
would be that in (3), the coefficient 50 would be replaced by m. A
candidate who tries to infer anything from the relationship between
50 and 100 will be only wasting his time. Maybe this was intended.
Sometimes, the ability not to get carried away by false clues is an asset.
3 3 3 3
Q. 35 If an = − ( )2 + ( )3 − . . . + (−1)n−1 ( )n and bn = 1 − an , then the
4 4 4 4
least natural number n0 such that bn > an for all n > n0 is ...... .

Answer and Comments: 5. This is a straightforward problem about

geometric progressions and inequalities, especially of powers. The pur-
pose of introducing bn is not clear. Since bn = 1 − an , the inequality
bn > an is equivalent to an < . The problem could have as well be
framed directly in terms of the latter inequality. Leaving this reduc-
tion to the candidate does not serve to test any great quality on his
part, except possibly the ability not to get bogged down by needlessly
clumsy formulation of a problem.
Now, coming to the solution, an is the sum of a G.P. with the first
3 3
term and common ratio − . Therefore by the formula for the sum
4 4

of the terms of a G.P., we have
− (− 43 )n )
an =
1 + 34
3 3
= 1 − (− )n (1)
7 4
3 3
Since | − | < 1, we know from (1) that an → as n → ∞. In
4 7
3 1
particular, since < , there will be some positive integer n0 such
7 2
that an < for all n > n0 . The problem asks us to find the least
integer n0 having this property.
We begin by recasting the inequality. Note that an < if and only
3 n 7
if 1 − (− 4 ) < , which in turn, is equivalent to
1 3 n
− < − (2)
6 4
The L.H.S. is negative while the R.H.S. is positive for even n. So (2)
holds for all even values of n. For odd values of n, say n = 2m + 1, (2)
1 3
reduces to − < − , or equivalently,
6 4
3 1
< (3)
4 6
The L.H.S. decreases as m increases because the base is less than 1.
In fact the L.H.S. tends to 0 as m → ∞. So, there is some value of m,
say m0 for which (3) holds. By monotonicity, it will then also hold for
all m ≥ m0 .
The least non-negative integer m for which (3) holds can be found
by taking logarithms w.r.t. some base greater than 1. Suppose we take
logarithms with base 10. Then keeping in mind that the logarithms of
both the sides are negative, we first reverse this inequality by taking
reciprocals before applying logarithms. Thus, (3) holds if and only if
>6 (4)

or equivalently,
log 6 log 6
2m + 1 > = (5)
log(4/3) log 4 − log 3

and finally, if and only if,

1 log 6 1 2 log 3 − log 2
m> ( − 1) = ( ) (6)
2 log 4 − log 3 2 2 log 2 − log 3

Thus the smallest integer m for which (3) holds is the smallest integer
which exceeds the R.H.S. of (6). But to evaluate this, we must be given
the values of the common logarithms of 2 and 3 (or logarithms w.r.t.
some other base). As these are not given, it is best to find the least m
for which (3) holds simply by trial. For m = 0, 1, 2 the L.H.S. of (3)
3 27 243 1
equals , , respectively. They are all bigger than . But the
4 64 1024 6
last one is already less than . So, for the next value of m, viz. m = 3
3 7 1 9 9 1
we have ( ) < × = < . Hence the smallest odd value of
4 4 16 64 6
n for which (2) holds is n = 7. But we already saw that it holds for
all even n anyway. Therefore if (2) is to hold for all n > n0 , then the
smallest such n0 is 5. If instead of n > n0 we were given n ≥ n0 , the
answer would be n0 = 6. (With a calculator, the numerical value of
the R.H.S. of (6) comes out to be 2.614131259. Hence the least integer
exceeding it is indeed 3 as we found by trial.)
The crux of the problem involves standard concepts and techniques.
The last part where we have to find the answer by trial is somewhat
unusual. When the answer involves logarithms and devices such as
log tables or calculators are not allowed, it is a standard practice to
leave the answer in a form like (6). (See for example, the JEE 2000
question, given in Comment No. 6 of Chapter 19.) Alternatively, the
statement of the question itself may give the relevant logarithms. In
the present problem, giving the values of log 2 and log 3 would have
enabled a candidate to get the answer from (6) without having to resort
to trial and error. But probably, most students would have found the
answer by trial anyway and since it is an objective type question, the
difference would never be known. If the requirement an < appearing

in the question (in a twisted from) had been replaced by something like
an < + 10−3 , then finding the least n by trial would have been too
time-consuming and the use of logarithms would have been mandatory.
Or, the data in the problem could have been retained as it is, but
instead of finding the integer n0 , the problem could have been asked in
a multiple choice format with the alternatives containing expressions
involving log 2 and log 3.

Q. 36 If f (x) is a twice differentiable function such that f (a) = 0, f (b) =

2, f (c) = −1, f (d) = 2 and f (e) = 0, where a < b < c < d < e, then
the minimum number of zeros of g(x) = (f ′ (x))2 + f ′′ (x)f (x) in the
interval [a, e] is ...... .

Answer and Comments: 6. We are not given the function f (x) ex-
plicitly. Nor are we given anything (such as a differential equation) from
which we can identify f (x) or g(x). If we could identify g(x) explic-
itly, then we can answer the question using the particular properties of
that function, just as we answer questions about zeros of trigonometric
functions by using various trigonometric identities.
Besides the existence of f ′′ (x), the only information we have about
f (x) is its values at the points a, b, c, d, e, which too are unknown except
for their relative order. So, this is an excellent example of a problem
where the lack of information is itself a clue. (See Comment No. 2 of
Chapter 24 for an elaboration of this apparently paradoxical situation.)
So, our only recourse is some theorems from calculus which assert
the existence of zeros of abstract functions under certain conditions
such as continuity and differentiability. At the JEE level, there are
only two such theorems. One is the Intermediate Value Property (IVP)
which says that if h(x) is continuous on an interval [p, q] and h(p), h(q)
are of opposite signs, then h(x) has at least one zero in (p, q). (See
Comment No. 3 for a proof and Comment No.s 4,5 and 6 for applica-
tions.) Another theorem, called Rolle’s theorem (Comment No. 8 of
Chapter 16) asserts that if h(x) is continuous on [p, q] and differentiable
on (p, q) and h(p) = h(q) = 0, then the derivative h′ (x) has at least
one zero in (p, q). A slightly improved version of this says that all we
need (besides the continuity and differentiability requirements) is that
h has equal values at the end-points, i.e. that h(p) = h(q) and not

that these values be 0. The improved version follows by applying the
earlier version to the function h(x) − h(p), or, as a very special case of
Lagrange’s Mean Value Theorem (MVT).
Let us now see which of these two results can help us in finding
the zeros of the function g(x) = (f ′ (x))2 + f ′′ (x)f (x). The IVP is not
of much help directly because although we are given the values of f (x)
at some points, from that we are not in a position to find the values of
g(x) (or even to determine their signs) at any point. Let us, therefore,
look at the only remaining hope, viz. Rolle’s theorem. For this we
must first recognise g(x) as a derivative of some function. And this is
indeed the case. In fact,
g(x) = (f ′ (x))2 + f ′′ (x)f (x) = (f (x)f ′ (x)) (1)
This recasting is the key to the solution. By Rolle’s theorem, between
any two zeros of f (x)f ′ (x) there is at least one zero of g(x). So the
problem is now reduced to finding the minimum number of zeros of the
function f (x)f ′ (x). Note that f (p)f ′(p) = 0 if and only if either f (p) =
0 or f ′ (p) = 0 (or both). So the problem further reduces to counting
the minimum numbers of zeros of each of the two functions f (x) and
f ′ (x). If we can show that they have, say, at least m and n zeros
respectively in [a, e], and further that none of these zeros is a common
zero of both, then it will follow that f (x)f ′ (x) has at least m + n zeros
and consequently that its derivative g(x) has at least m + n − 1 zeros.
Let us tackle these two tasks separately. First the zeros of f (x) in
[a, e]. We are already given that f (a) = 0 and f (e) = 0. So, these are
two zeros of f (x). The remaining zeros have to come from the IVP.
Since f (b) = 2 > 0 and f (c) = −1 < 0, there exists z1 ∈ (b, c) such
that f (z1 ) = 0. Similarly, from f (c) = −1 < 0 and f (d) = 2 > 0, there
exists z2 ∈ (c, d) such that f (z2 ) = 0. Further,
a < b < z1 < c < z2 < d < e (2)
So these four zeros of f (z), viz. a, z1 , z2 , e are distinct. By Rolle’s
theorem between every consecutive two of them, there is at least one
zero of f ′ (x). Hence there exist z3 ∈ (a, z1 ), z4 ∈ (z1 , z2 ) and z5 ∈ (z2 , e)
such that f ′ (z3 ) = f ′ (z4 ) = f ′ (z5 ) = 0. Further,
a < z3 < z1 < z4 < z2 < z5 < e (3)

Thus we have found at least seven distinct zeros of f (x)f ′ (x). By
Rolle’s theorem again, its derivative, which is g(x) has at least 6 zeros
in (a, e).
Strictly speaking, the solution is not yet complete. We have shown
by some argument that if f (x) satisfies the conditions in the problem,
then the corresponding function g(x) = (f ′ (x))2 + f (x)f ′′ (x) has at
least 6 zeros in (a, e). But conceivably, using some other argument we
(or somebody else) can show that under the same conditions g(x) must
have at least 7 distinct zeros. The problem demands that we show that
this is impossible, or in other words, that in general 6 is the best lower
bound on the number of zeros of g(x). To do this it suffices to give
an example of a function f (x) which satisfies all the conditions of the
problem and has exactly 6 and no more zeros. The graph of one such
function is shown below. Note that here z3 , z4 , z5 coincide with b, c, d

a . . .c . .
b z1 z2 d e

Constructing such a function by an explicit formula is a little

messy. If instead of f (c) = −1 we had f (c) = −2, then the function
f (x) = 2 sin x with a = 0, b = π/2, c = 3π/2, d = 5π/2 and e = 3π
would work. Here g(x) comes out to be (4 sin x cos x) = 4 cos 2x and
it has exactly 6 zeros in the interval [0, 3π], viz. all odd multiples of
π 11π
from to .
4 4
Unlike the last three questions, here the examiners have attempted
to pose a question of theoretical calculus as a fill in the blank question.
There are some inherent limitations on the success of such attempts.
The real crux of such problems is the arguments that are needed in
justifying the steps rather than some number associated incidentally

with the problem. When the correct answer is a four digit number like
that of Q. 33 or Q. 34, a correct answer is unlikely to be arrived at by
wrong reasoning. But that is not the case for a problem like the present
one where the answer is a small one, viz. 6. The figure 6 is easy to
come up with by a totally wrong argument. For example, a candidate
may think that since f (x) has two given zeros, so must be the case with
f ′ (x) and also with f ′′ (x). Therefore the function (f ′ (x))2 + f (x)f ′′ (x)
has 6 zeros since it involves all three, viz. f (x), f ′(x) and f ′′ (x).
But the real loss is that a candidate who has the fineness of
realising that the solution is not over until 6 is shown to be the best
answer stands to get no reward. Similarly, as commented earlier, even
though Q. 33 was purely computational, a candidate who scrupulously
dismisses the degenerate cases there is only wasting his time.

Fill-in-the-blank type questions were asked in the JEE long time ago.
That time they were evaluated manually. As a result, the examiner could
adjudicate whether to allow sin π2 as a correct answer even though it was not
simplified to 1. But then controversies would arise as to what degree of sim-
plification is expected from the candidates. For example, in a combinatorial  
or probability problem, it is reasonable to expect that something like 93 be
simplified to 84 and  equally reasonable not to expectany  simplification for
something like 365 20
. But what about something like 13
Asking that the answer be a whole number eliminates such controversies.
This, in fact, is close to the style adopted for the answers of the Main Problem
in each chapter of Eductive JEE Mathematics. The merits and demerits of
this practice are discussed in Comment No. 1 of Chapter 2. The major
demerit is that a candidate who has correctly done all the work except for a
minor numerical slip loses heavily as there is little to warn him of his mistake.
For the past several years, the objective part (the Screening Paper) con-
sisted wholly of multiple choice questions. But some questions are simply
not suited to be asked in this format. So they had to be asked only in the
Main Paper, where the candidates would have to show the reasoning. Now
the second round is eliminated completely. Apparently, the intention is to
use the fill in the blank type questions to fill the gap thus created. But out
of the four questions in this Section only one (viz. Q. 34) can be said to fully
meet the demand.

Match the entries given in the two columns in
each question. One entry in the first column may have
more than one matching in the second and vice versa.

Q. 37 Normals are drawn at points P, Q and R lying on the parabola y 2 = 4x

which intersect at (3, 0). Then
(i) Area of ∆P QR (A) 2
(ii) Radius of circumcircle of ∆P QR (B) 5/2
(iii) Centroid of ∆P QR (C) (5/2, 0)
(iv) Circumcentre of ∆P QR (D) (2/3, 0)

Answer: (i) ↔ (A), (ii) ↔ (B), (iii) ↔ (D), (iv) ↔ (C).

Comments: It is customary (and logical) to specify areas in terms
of square units rather than a mere number. This convention has been
followed in the wording of Q. 29. Had it been followed in this question
too while giving the options in the second column, then the matching
option for (i) in the left column could have been found without doing
any work, because it is the only option which is expressed in terms
of sq. units. That would be a superb tribute to the ability to get at
the right answer by merely looking at the format of the answer, much
the same way that a Sherlock Holmes can identify a murderer correctly
simply by looking at the clothes of the suspects.
It is perhaps to avoid this unwarranted short cut that the paper-
setters have given the area as a pure number. But even as the question
stands, there is ample room for sneaking in. The first column lists
four geometric attributes of the triangle P QR. Two of these ((iii) and
(iv)), are certain points, viz. the centroid and the circumcentre of the
triangle. When they are specified by coordinates, the corresponding
entries in the right column have to be ordered pairs of real numbers.
There are only two such entries, viz. (C) and (D). So, by sheer common
sense, one of these two has to match (iii). Moreover, the centroid and
the circumcentre coincide only for an equilateral triangle. So, if we can

eliminate this possibility, which is often easy to do, we automatically
get the match of (iv) as soon as we have found that of (iii). Also, if we
can tell by inspection that one of (C) and (D) is outside the triangle,
then it cannot possibly be the centroid.
In fact, in the traditional form of questions asking for matching
the pairs, it was inherent that the matching be a bijective one. And
the charm of such questions was that after making all but one matches
correctly, the last one came in as a bonus. Apparently, the paper-
setters have decided to disallow this bonus. That is why it is given
that the same entry in the first column may have several matches in
the second. This effectively means that in the first column, we are
having four separate problems, each with one or more correct answers
in the second column. And, we have to tackle them one-by-one.
A typical point on the parabola y 2 = 4x is of the form (t2 , 2t) for
some value of the real parameter t. The slope of the tangent at this
point is and so that of the normal is −t. Therefore the equation of
the normal at this point is

y − 2t = −t(x − t2 ) = t3 − xt (1)

Let the three points P, Q, R on the parabola correspond to t1 , t2 and

t3 respectively. Then for these values of t, the normal is given to pass
through the point (3, 0). Combined with (1), this means that t1 , t2 , t3
are the roots of the cubic equation

t3 − t = 0 (2)

(Points on a parabola the normals at which are concurrent are called

co-normal points. So the present problem deals with a triangle whose
vertices are co-normal points on a given parabola. We could also have
started with the equation of the normal to a parabola y 2 = 4ax in
terms of its slope m, viz. y = mx − 2am + am3 . In that case, instead
of (2) we would have gotten the cubic equation

m3 − m = 0 (3)

But thereafter we would have to express the points P, Q, R in terms of

m, for which we would have to express m in terms of the parameter

t anyway. And that would take us to (2). The point is that such
specialised formulas as the equation of a chord in terms of its midpoint,
or the equation of a normal in terms of its slope, should be applied with
discretion. Sometimes they do save precious time. But they are not
golden tools all the time.)
Although a general cubic equation is not very easy to solve, (2) is
an exception. By inspection, its roots are 0, 1 and −1. As a result, we
can take P = (0, 0), Q = (1, 2) and R = (1, −2). The parabolical part
of the problem is now over. From now onwards it is a simple problem
in coordinate geometry. Taking the A.M. of the x-coordinates and
also that of the y-coordinates of the vertices, we immediately get that
the centroid of ∆P QR is at (2/3, 0). As noted before, this leaves the
circumcentre no other choice but (5/2, 0). Still, to get this honestly, we
note that the x-axis is the perpendicular bisector of the segment QR.
So, the circumcentre must be of the form (h, 0) for some h. Equating
its distances with the three vertices we get h2 = (h − 1)2 + 4 which
gives h = 5/2. Honesty has paid, because the circumradius is now h
which is 5/2. As for the area, ∆P QR is isosceles with base QR = 4
and altitude 1. So the area is 2 sq. units.
Basically, this is a trivial problem. It would have been a good prob-
lem if instead of (2), we ran into a cubic equation which was not easy to
t2 + t22 + t23 2(t1 + t2 + t3 )
solve. In that case, the centroid would have been ( 1 , ).
3 3
As the expressions in the numerators are symmetric functions of the
roots t1 , t2 , t3 , they can be evaluated even without solving the cubic.
(See Exercise (9.61) for a far more challenging problem of this spirit.
Gone are the days when gems like this were asked in the JEE.)
Another criticism of this problem is that in the given triangle,
side QR is the latus rectum of the parabola and the third vertex P is
also the vertex of the parabola. This was also exactly the case in Q. 29
above, where too, the area of the triangle was asked. This duplication
in the same paper is shocking to say the least. And to make it worse,
Q. 29 itself has some overlap with Q. 10 as remarked there. So what
we have is not a duplication but a triplication of ideas!

Q. 38 Match the following:

(i) (sin x)cos x (cos x cot x − ln[(sin x)sin x ]dx (A) 1
(ii) Area bounded by −4y 2 = x and x − 1 = −5y 2 (B) 0
(iii) Cosine of the angle of intersection of the curves (C) 6 ln 2
y = 3x−1 ln x and y = xx − 1
(iv) Some problem about the differential equation (D) 4/3
dy 2
= with y(1) = 0
dx x+y

Answer: (i) ↔ (A), (ii) ↔ (D), (iii) ↔ (A), (iv) ↔ (uncertain)

Comments: Unlike in the last question where all the four entries
in the first column had a common setting, here it is a bunch of four
totally unrelated problems. The details of the problem in (iv) were
irretrievable. Had this happened in a question of matching the pairs in
the traditional form, it would not have mattered. Once the other three
entries in the first column are found matches, the match of the last one
is fixed in heaven no matter what the problem is. That is not the case
here. So, let us tackle the first three problems fully and the fourth one
to the extent we can.
The integrand in (i) looks horrendous because of the exponen-
tials in it with bases other than e. Let us convert it to an expo-
nential with base e so that it looks less formidable. Since sin x =
eln sin x , we get (sin x)cos x = ecos x ln sin x . If we differentiate this, we get
ecos x ln sin x (cos x ln sin x) = ecos x ln sin x (cos x cot x−sin x ln sin x). But
if we convert the two powers to the base sin x, we get
((sin x)cos x ) = (sin x)cos x (cos x cot x − ln[(sin x)sin x ]) (1)
But the R.H.S. is precisely the integrand in (i). So we are lucky. The
integral then equals (sin x)cos x = 10 − 01 = 1.

(ii) is a typical problem of evaluating an area bounded by two curves.
Both the curves are parabolas. Going by the number of problems they
have asked, it looks like the paper-setters this year are especially fond
of parabolas! The two given parabolas meet when −4y 2 = 1 − 5y 2 ,
which gives y = ±1 and x = 4. So the region bounded by them is as
shown in the figure below. Clearly it is symmetric about the x-axis

since both the curves are so. Therefore it suffices to find the area of
the upper half. It is more convenient to do this by horizontal, rather
than vertical slicing.

(− 4, 1)

x = − 4y2 x = − 5 y2 + 1
O (1, 0)

(− 4, − 1)

By horizontal slicing as shown the desired area, say A, equals

Z 1
A = 2 (−5y 2 + 1) − (−4y 2)dy
Z 1
= 2 1 − y 2 dy
= 2(y − y 3/3) = 4/3 (2)


As compared to some area problems in the past, the one this year
is quite straightforward. But in the past, such questions would get
about 5 to 8 minutes. For the present problem, the time allowed is less
than one minute! Once again, it is no consolation that only the final
answer needs to be shown. There is no smart way of simply guessing
the answer. To arrive at it, you have to do a lot of work anyway,
whether you show it or not.
In the case of Item (iii), by inspection we can identify (1, 0) as a
point of intersection of the two curves y = 3x−1 ln x and y = xx − 1.
The angle of intersection of the two curves is the angle between their
tangents at the point of intersection. Now, by direct calculation,

d x−1 3x−1
(3 ln x) = + 3x−1 (x − 1) ln 3 ln x (3)
dx x

whose value at x = 1 is 1. Hence the slope of the tangent to the first
curve at the point (1, 0) is 1. Similarly,

d x d x ln x
(x − 1) = (e − 1)
dx dx
= xx ( (x ln x))
= xx (1 + ln x) (4)

whose value at x = 1 is also 1. Therefore the two curves have the same
tangents at their point of intersection. Hence the angle between them
is 0 and and its cosine is 1. Note that both the entries (i) and (iii) have
the same match, viz. (A) in the second column, even though they are
totally unrelated to each other.
Strictly speaking, the solution is not complete. We have found one
point of intersection of the two curves by inspection. But there could
be others which are not easy to find. (Occasionally, the paper-setters
do miss some possible solutions, thereby making the problem almost
impossible to solve. See the JEE 1996 and JEE 1987 problems discussed
in Comment No. 14 of Chapter 10.) Moreover, in the present problem,
the angles of intersection of the two curves at these other points may
be different than the one at (1, 0). As the question is framed to have a
single answer, we simply have to assume that there are no other points
of intersection. A candidate who tries to find them or to prove their
non-existence is wasting his time. Yet another instance where scruples
do not pay.
Finally, let us tackle Item (iv) to the extent we can by solving the
differential equation given, viz.
dy 2
= (5)
dx x+y
with the initial condition that y = 0 when x = 1.
As it stands (5) cannot be cast in the separate variables form. Nor
does a substitution like y = vx work as one sees by trial. But there is
x+y 2
a trick. If the R.H.S. were instead of , then the equation
2 x+y
would have been a linear first order d.e. which can be solved by the

method given in Comment No. 12 of Chapter 20. That suggests the
trick. Why don’t we interchange the roles of x and y? Normally, we
take x as the independent variable and y as the dependent variable,
i.e. a function of x. But there is nothing sacrosanct about this if we
are only looking for solution curves which are of the form f (x, y) = c
where the roles of the two variables are on par. So, we take reciprocals
of both the sides of (5) and get

dx x+y
= (6)
dy 2

(In JEE 2005 too, the Screening Paper had one d.e. which was amenable
to the very same trick. See Q. 25 in the Educative Commentary on JEE
2005 by the author. But that time an alternate approach was also pos-
sible. That does not seem possible with the present problem.)
Rewriting (6) as

dx x y
− = (7)
dy 2 2

we get e−y/2 as an integrating factor (see Comment No. 12 of Chapter

19) and

x = −y − 2 + cey/2 (8)

as the general solution, where c is an arbitrary constant. The initial

condition y = 0 when x = 1 determines c as 3. Therefore the particular
solution of the d.e. in Item (iv) is

x + y + 2 = 3ey/2 (9)

We cannot proceed further because it is not known what is asked.

Q. 39 Match the following.

(i) Two rays x + y = |a| and ax − y = 1 (A) 2
intersect each other at a point in the
first quadrant for all a in the interval
(a0 , ∞). The largest value of such a0 is
(ii) Point (α, β, γ) lies on the plane (B) 1
x + y + z = 2. If ~a = αî + β ĵ + γ k̂
a) = ~0, then γ equals
and k̂ × (k̂ × ~
R1 R0 R1 √
(iii) (1 − y 2 )dy + (y 2 − 1)dy (C) 1−x dx

0 1 0

R0 √

+ 1+x dx


(iv) If sin A sin B sin C + cos A cos B = 1, (D) 4/3

then sin C equals

Answer: (i) ↔ (B), (ii) ↔ (A), (iii) ↔ (C) and (D), (iv) ↔ (B).
Comments: Another bunch of four totally unrelated problems. But
there is even more. The normal practice in a match the pairs question
is that the first column contains some problems and the second one
contains their (usually numerical) answers. But in this question item
(C) in the second column is an integral. When evaluated, it could as
well come out to be the value of sin C in item (iv) in the first column.
We would not know until we find both. So, in effect, in this single
question there are five different problems, to be solved for 6 points. In
terms of proportional time this means less than four minutes! In effect,
then, each one of these five questions is to be understood and worked
out in less than 48 seconds. Moreover, even though the problems are
totally unrelated to each other, to claim the 6 marks for the question,
all have to be answered correctly. This is almost like saying that in
order to enter the I.I.T.’s you must not only do well in the JEE, but
simultaneously you (or perhaps your sister) must also qualify for a
medical admission!
Anyway, getting down to business, let us tackle these five problems,
one-by-one. The word ‘ray’ in the statement of (i) is a little misleading.
No harm would arise if it is replaced by ‘line’, or better still, by ‘straight
line’. A ‘ray’ is a half-line, i.e. a portion of a straight line lying entirely

on one side of some point P on it. (The point P is then called the
initial point of the ray.) In the present instance, the portion of the line
x + y = |a| in the first quadrant is not a ray but a line segment with
end points (|a|, 0) and (0, |a|).
Assuming, therefore, that the word ‘ray’ simply means a ‘line’ we
solve the two equations simultaneously to get the coordinates of their
point of intersection as

|a| + 1
x = (1)
a|a| − 1
and y = (2)

Since the point of intersection lies in the first quadrant, we have x > 0
and y > 0. As the numerator of x is always positive, we must have
a + 1 > 0, which gives a > −1 or equivalently a lies in the interval
(−1, ∞). But we must not hastily conclude a0 = −1, for, the positivity
of y may restrict a still further. This indeed happens, because as the
denominator of y is now known to be positive, we must also have a|a| >
1. This automatically rules out negative values of a and thereby implies
a|a| = a2 . So the positivity of the numerator of y now implies a > 1,
i.e. a ∈ (1, ∞). Hence we must take a0 as 1. (As in Q. 36, in a
problem like this a complete solution does not end here. The interval
of values which a could assume under the conditions of the problem
was already narrowed down from (−1, ∞) to (1, ∞). Strictly speaking,
we must now show that no further narrowing is possible. That is, for
every a ∈ (1, ∞), we must show that the two given lines intersect at
a point in the positive quadrant. In the present case, no further work
is needed for this, because for every a ∈ (1, ∞), (1) and (2) show that
x, y are both positive.)
Unlike the entry (i) in the first column, entry (ii) is quite clear and
straightforward. The first condition simply means

α+β+γ =2 (3)

(It is really difficult to see what is gained by giving this piece of data in
such a twisted form. It does not test intelligence. It only slows down

an intelligent student and thereby becomes more a test of his speed.)
As for the vector triple product, we have, by a standard identity,

k̂ × (k̂ × ~a) = (k̂ · ~a)k̂ − (k̂ · k̂)~a (4)

Since ~a = αî + β ĵ + γ k̂ we have k̂ · ~a = γ. Also k̂ · k̂ = 1. Putting these

values in (4),

k̂ × (k̂ × ~a) = γ k̂ − ~a = γ k̂ − (αî + β ĵ + γ k̂) = −αî − β ĵ (5)

As we are given that k̂ ×(k̂ ×~a) = ~0, we now get α = 0 = β. Combining

this with (2) we get γ = 2.
Entry (iii) in the first column contains two very easy integrals.
Evaluating them, we get
1 0
y 3 1 y 3
(1 − y 2 )dy + (y 2

− 1)dy = (y − ) + ( − y)

3 0 3 1
0 1
2 2 4
= + = (6)
3 3 3
Thus we get that (iii) matches with (D). Instead of going to Item (iv),
let us first finish off with the two integrals in (C). Calculation of these
two integrals is almost as easy as the ones in (iii). As both the inte-
grands are positive (and the lower limits are smaller than the upper
ones) the absolute value signs are redundant. Therefore,
1 0
Z1 √ Z0

√ √ √

1 − xdx +
1 + xdx
= 1 − xdx + 1 + xdx

0 −1 0 −1
2 1 2 0
= (− (1 − x)3/2 ) + ( (1 + x)3/2

3 0 3 −1
2 2 4
= + = (7)
3 3 3
So, we see that (iii) and (C) also match each other. We did this by
actually computing both of them. And since all the four integrals are
easy to evaluate, any improvement seems pointless. The problem would
have been more interesting if the integrals on each side did not permit
easy evaluations individually but their sum was easy to evaluate. And

the problem would be still more interesting if the integrals were such
that their evaluations were not easy but by suitable substitutions the
integrals in (C) could be transformed to those in (iii). But all this can
hardly be expected in a problem which is to be solved in 48 seconds.
We now come to the last part, viz. (iv) in the first column. We
are given an equation

sin A sin B sin C + cos A cos B = 1 (8)

and are asked to determine the value of sin C. Most people would
instinctively take A, B, C to be the angles of a triangle. But this ought
to have been specified in the statement of the problem. Without this
hypothesis, sin C cannot be determined uniquely. Suppose, for example
that A = B = 0. Then (8) holds for every value of C and hence
nothing can be said about the value of sin C (except, of course, that
it lies in [−1, 1]). So, we assume that that A, B, C are the angles of a
triangle. In that case, the problem is a virtual replication of a 1986 JEE
problem, solved in Comment No. 10 of Chapter 14 on trigonometric
optimisation. Even the notations have not been changed! (In terms of
time allowed, the difference is staggering. In 1986, you were given the
answer. Specifically, √you were asked to show (with reasoning) that the
a : b : c = 1 : 1 : 2 and you got 9 minutes for it. Now you get 48
seconds to come up with sin C = 1.)
Normally, when we have three unknowns A, B and C we need
three equations to determine them. One of these is (8) while the other
is given by our assumption, viz. A + B + C = π. In general two
equations cannot determine the unknowns uniquely. The only way this
can happen is that one side of the equation represents the optimum
value of the other. As a simple example, if α, β lie in, say [0, π], then
from the value of cos α + cos β we cannot determine them uniquely.
But if this value happens to be 2, then it is the maximum value of
cos α + cos β and it can occur only when both cos α and cos β both
equal 1 each. That determines α, β.
We apply a similar reasoning here. Since sin A, sin B are non-
negative and sin C ≤ 1, the L.H.S. of (8) is at most sin A sin B +
cos A cos B which is simply cos(A − B), which can never exceed 1. So
(8) forces the equality of cos(A − B) and 1. The only way this can

happen is if A = B. Therefore, (8) now becomes

sin2 A sin C + cos2 A = 1 (9)

1 − cos2 A
From this we see that sin C = = 1.
sin2 A
Q. 40 Match the following:

(i) If = t, then tan t equals (A) 2 2
i=1 2i2
(ii) Sides a, b, c of a triangle ABC are in A.P. (B) 1
a b
and cos θ1 = , cos θ2 = and
b+c c+a
c θ1 θ3
cos θ3 = , then tan2 + tan2 equals
a+b 2 2 √
(iii) a line is perpendicular to x + 2y + 2z = 0 and (C)
passes through (0, 1, 0). Then the perpendicular
distance of this line from the origin equals
(iv) A plane passes through (1, −2, 1) and is (D) 2/3
perpendicular to the two planes 2x − 2y + z = 0
and x − y + 2z = 4. Then the distance of the
plane from the point (1, 2, 2) is

Answer: (i) ↔ (B), (ii) ↔ (D), (iii) ↔ (C), (iv) ↔ (A).

Comments: Entry (i) is a direct continuation of Exercise (10.20),
which asks for the partial sum, say Sn of the given series. In other
1 1 1 1
tan−1 ( ) = tan−1 ( ) + tan−1 ( ) + . . . + tan−1 ( 2 )
Sn = 2
i=1 2i 2 8 2n

There is no standard or obvious formula for this sum. But we can learn
a little by experimenting with a few small values of n. The well-known
tan A + tan B
identity tan(A + B) = translates into
1 − tan A tan B
tan−1 (x) + tan−1 (y) = (2)
1 − xy

which is valid at least when both x, y lie in [0, 1). Repeated applications
of this give,
S1 = tan−1 ( ) (3)
1 1 1/2 + 1/8 2
S2 = tan−1 ( ) + tan−1 ( ) = tan−1 ( ) = tan−1 ( ) (4)
2 8 1 − 1/16 3
2 1 2/3 + 1/18 39
S3 = tan−1 ( ) + tan−1 ( ) = tan−1 ( ) = tan−1 ( )
3 18 1 − 1/27 52
= tan−1 ( ) (5)
The pattern is now clear. For every n ∈ IN , we have
Sn = tan−1 ( ) (6)
This can be proved by induction on n using (2). But in an objective
type test that hardly matters. By taking the limit as n → ∞, we get

tan−1 (
t= ) = lim Sn
i=1 2i2 n→∞

= lim tan−1 ( )
n→∞ n+1
= tan−1 ( lim ) (7)
n→∞ n + 1

= tan−1 (1)
= (8)
where in (7) we have used the continuity of the arctan function at 1
and Theorem 1 of Comment No. 3 of Chapter 16. So, t = π/4 and
hence tan t = 1.
In a conventional examination, a proof of (6) would have to be
given. With it this would be a good problem, especially because it tests
the ability to make experiments, recognise some pattern, come up with
a guess and finally to prove the guess.
Item (ii) is also a good problem in trigonometry. But its beauty
is marred by the objective type format of the test, exactly the same
way as that of Q. 27. The real crux of the problem is to show that if

θ1 θ3
the sides a, b, c are in an A.P. then tan2 ( ) + tan2 ( ) is a constant.
2 2
Finding the value of this constant is a relatively minor matter. But in
an objective test, this minor matter is all that matters. So, to find the
value of this expression, all that we have to do is to take a particular
θ1 θ3
triangle ABC of our choice and compute tan2 ( ) + tan2 ( ) directly
2 2
for it. All we have to ensure is that a, b, c are in an A.P. We could,
for example, take ABC to be a right angled triangle with a = 3, b = 4
and c = 5. But nothing can beat the case of an equilateral triangle.
This is perfectly legitimate because it is nowhere required that three
numbers in an A.P. have to be distinct. So, taking a = b = c, we have
cos θ1 = cos θ2 = cos θ3 = . Therefore, θ1 = θ2 = θ3 = 60◦ . Hence
2 θ1 2 θ3 2
tan ( ) + tan ( ) = 2 tan2 (30◦ ) = . The answer is complete!
2 2 3
As in the case of Q. 27, maybe the paper-setters intended this
short cut because that is all you can reasonably expect in a minute.
Still, for the sake of completeness, we give a proof of the constancy of
θ1 θ3
tan2 ( ) + tan2 ( ) under the conditions of the problem. We begin by
2 2
expressing tan2 ( θ2 ) in terms of cos θ by

θ 2 sin2 ( θ2 ) 1 − cos θ
tan2 ( ) = 2 θ = (9)
2 2 cos ( 2 ) 1 + cos θ

Applying this for the given θ1 and θ3 and adding

θ1 θ3 1 − cos θ1 1 − cos θ3
tan2 ( ) + tan2 ( ) = +
2 2 1 + cos θ1 1 + cos θ3
b+c−a a+b−c
= +
b+c+a a+b+c
2b 2b 2
= = = (10)
a+b+c 3b 3
since a + c = 2b. So, an honest proof was not, after all, difficult. But
when you have to get it in less than a minute, dishonesty is tempting.
Item (iii) deals with the distance of a point from a line. In Q.
10 too, we encountered the distance of a point from a line. But that
time both were in the xy-plane. Now they are in the three dimensional

space. So this problem is qualitatively different. Let L be the given
line. Since it is perpendicular to the plane x + 2y + 2z = 0, it is parallel
to the normal to this plane, and hence to the vector î+ 2ĵ + 2k̂. Since it
passes through (0, 1, 0) its parametric equation in the vector form can
be written down as

~r = ~r(t) = tî + (1 + 2t)ĵ + 2tk̂ (11)

Let the perpendicular from the origin O to L fall at the point P = ~r(t0 ).
Then OP ⊥ L which gives

~r(t0 ) · (î + 2ĵ + 2k̂) = 0 (12)

(11) and (12) together give t0 + 2(1 + 2t0) + 4t0 = 0 i.e. t0 = − , which
2 5 4
determines P as (− , , − ). Therefore the distance of O from L, i.e.
9 9 √9 √
−→ 4 + 25 + 16 5
the length of OP , equals = .
9 3
Item (iv) asks for the distance of a point from a plane. Since the
plane passes through (1, −2, 1), its equation can be written in the form

a(x − 1) + b(y + 2) + c(z − 1) = 0 (13)

where a, b, c are some constants to be determined. The perpendicu-

larity conditions give the following system of two homogeneous, linear
equations in the three unknowns a, b, c.

2a − 2b + c = 0 (14)
and a − b + 2c = 0 (15)

This system has no unique solution. But by Theorem 7 in Com-

ment No. 17 in Chapter 3, it determines the relative proportions of
a, b, c. Specifically,
we get that
a, b, c are proportional to the num-
−2 1 1 2 2 −2
bers , and respectively. So, we may take

−1 2 2 1 1 −1
a = −3, b = −3 and c = 0, or still better, a = 1, b = 1 and c = 0.
Putting these in (13), we get the equation of the plane as

x+y+1= 0 (16)

The distance of the point (1, 2, 2) from this plane is √ 2 =
√ 1 + 12 + 02
2 2.
Essentially the same solution can be presented a little differently
using the concept of cross product. Since the plane is perpendicular to
the plane 2x−2y +z = 0, it is parallel to the normal to the latter plane.
Thus we get that the given plane is parallel to the vector 2î − 2ĵ + k̂.
Call this vector ~u. Similarly the perpendicularity of the plane with the
plane x − y + 2z = 0 makes it parallel to the vector î − ĵ + 2k̂ (=
~v) (say). Therefore the plane is perpendicular to the cross product of
these two vectors, viz. to

î ĵ k̂

~u × ~v = 2 −2 1

1 −1 2

= −3î − 3ĵ (17)

So, this vector is perpendicular to the plane. Further the plane passes
through the point (1, −2, 1). So its equation is

−3(x − 1) − 3(y + 2) + 0(z − 1) = 0 (18)

which reduces to (16). The rest of the work is the same.

When approached in this manner, the problem overlaps consid-
erably with Q. 20 above, where too, the key idea was that a vector
perpendicular to each of two given vectors is parallel to their cross
product. Yet another duplication, which seems to have no purpose.
Whichever method is used in the solution, the problem is extremely
straightforward. The formula for the distance of a point from a plane
is the direct analogue of that for the distance of a point from a line in a
plane. The latter was already used in the solution to the problem above
about a parabola (Q. 10). This duplication could have been avoided.
Apparently it is necessitated because in order to bring Item (iv) on par
with the other three items in the first column, its answer has to be a
single real number. Had it been left at merely asking for the equation
of the plane, then the answer in the second column would have been
x + y + 1 = 0 and it would have been obvious that this cannot be the
answer to any of the other three items. So this is another example

how the inherent constraints of formatting compel the paper-setters
to append a problem with relatively unimportant ancillaries, thereby
increasing the possibility that a candidate who has got the conceptual
part of the problem correctly gets a rude jolt because of a silly slip in
handling these appendages.

The year 2006 is remarkable in the long history of the Joint Entrance Ex-
amination. Right from its start more than forty years ago, the JEE remained
a conventional type examination till 1978. That is, you had to show not only
the answers but the working for them.
The objective type questions made their first appearance in 1978. Their
nature, purpose and importance varied considerably over the years. For
example, initially the multiple choice questions had only one correct answer.
Later on MCQs with one or more correct answers were introduced. Negative
marking was introduced, then withdrawn for many years and re-introduced
from 2002 onwards. Mechanical evaluation was started a few years earlier.
From 2000 to 2005, the JEE was conducted in two rounds. The first or the
Screening round was fully objective. Only those who cleared it were allowed
to appear for the second or the Main round. The papers in the Main round
were the conventional types and the final selection was made solely on the
basis of the performance in the second round. In effect, this meant that the
screening was done efficiently with objective type questions but the ultimate
selection was based solely on conventional types of questions.
It is for the first time in 2006 that the selection is based entirely on a
single, fully objective type paper in each subject. This unprecedentedness
warrants a more quantitative analysis of the question paper than we have
been doing for the past three years (2003 to 2005) in the respective educative
In all, the Mathematics question paper, to be completed in two hours,
carries 184 marks. (On a proportionate time scale, this means you get less
than 40 seconds per mark.) Only 24 marks out of these (Q. 33 to Q. 36)
are for the fill in the blank questions, where the answers are to be filled in
by circling the appropriate digits of a four digit number. The remaining 160
marks are for multiple choice questions. The last four questions (Q. 37 to Q.
40) ask for matching entries in one column with those in the second. But the
same entry is allowed to have several matches. As a result, even if you know
the matches of all but one, you do not automatically know the match of the
last one. In effect this means that each entry is like an independent multiple
choice question with one or more correct answers from the given ones. Also

as in Q. 39, occasionally even the second column contained an item which
first had to be evaluated before deciding if it matched with any entry in the
first column.
Counting all these possibilities, the present JEE 2006 Mathematics ques-
tion paper has 53 different problems to be worked out in two hours. The
following tables give a classification of these problems. Listed against each
problem are its location (L) i.e. its question number (as listed in this com-
mentary), the number of marks (M) it carries and the broad area(s) of mathe-
matics it comes from. This is followed by seven columns headed by the letters

T, C, S, I, U, F and R. A tick mark ( ) under these columns symbolises the
following qualities:

T : (Thought oriented). This means that the success requires a certain

key idea which will come only with a correct line of thinking.

C : (Computational). This means that the problem involves a fair amount

of computation which can go wrong because of silly slips.

S : (Sneaky approach). This means that there is an efficient but wrong

way to get to the right answer, or a part of it which gets rewarded
because the candidate does not have to show the reasoning. (Not to
be confused with an elegant solution, which is perfectly legitimate, and
comes under T.)

I : (Improper). This means there is some impropriety such as some am-

biguity or redundancy (not necessarily a mathematical mistake) or a
poor notation or clumsiness in the statement of the problem or annex-
ation of tidbits irrelevant to the main theme of the problem.

U : (Unscrupulous). This means that during or even after getting the

correct answer, some additional justification will have to be given to
complete the solution scrupulously had the same problem been asked
in a conventional type examination.

F : (Familiar problem). This means a substantially similar problem has

been asked in the past or is included in Educative JEE Mathematics
by the author. (The exact reference is given in the comments on that

R : (Repetitious problem). This means that the work needed overlaps
considerably (except for numerical differences) with another problem
in this paper (whose serial number appears in the column).

Needless to say that such categorisation is open to controversies. Many

problems require a combination of thought and computation and sometimes
it is a matter of opinion which of the two dominates. If the thought needed is
a commonplace (e.g. that of finding a vector perpendicular to a given plane),
then the problem is not classified under T. In some problems both a thought
and a computation are needed strongly and so the tick mark occurs in both
the columns. When an elegant approach is possible, the problem is classified
under T. Similarly whether a problem falls under U or not depends on the
degree of expected completeness. The standards adopted here are those in the
author’s Educative JEE Mathematics. Also, some problems not listed under
F here may be familiar because of inclusion in some other commonly used
source. As for the area covered by a problem, sometimes it is a combination
of two areas. But when one of them is superficial, we do not list it.

S. Q. M Area(s) T C S I U F R
No. No.

1 1 3 Trigonometry, inequalities

2 2 3 Evaluation of Limits
√ √
3 3 3 Solution of triangles √ √ √
4 4 3 Trigonometric optimisation

S. Q. M Area(s) T C S I U F R
No. No.
√ √
5 5 3 Quadratic inequalities,
trigonometric equations
√ √ √ √ √
6 6 3 Complex numbers √ √
7 7 3 Number theory,
combinatorics √ √ √
8 8 3 Finding antiderivatives
√ √ √ √
9 9 3 Differential equations
√ √ √
10 10 3 Equation of parabola √ √ √ 53
11 11 3 Integrals,

12 12 3 Vector projection √ √ √ 20
13 13 5 Tangents to parabola
√ √
14 14 5 Ellipse, hyperbola √ √
15 15 5 Solution of triangles
√ √
16 16 5 Continuity, differentiability
√ √ √
17 17 5 Finding a cubic √
18 18 5 Skew-symmetric matrix
√ √ √ √
19 19 5 Differential equations

20 20 5 Angle between 12,
two vectors 53
√ √ √
21 21 5 Probability, limits √
22 22 5 Conditional probability 23

23 23 5 Conditional probability 22
√ √
24 24 5 Approximate integration √ √ √
25 25 5 Maxima/minima
√ √ √
26 26 5 Theoretical calculus √ √ √
27 27 5 Incircle and circumcircle
of a square
√ √
28 28 5 Identifying locus type √ √
29 29 5 Triangle inscribed 10,
in a parabola 37

30 30 5 Determinant of a matrix √ √
31 31 5 Sum of matrix entries

32 32 5 Product of matrices √ √ √
33 33 6 Quadratic equations
√ √ √
34 34 6 Reduction formula for
definite integrals √
35 35 6 G.P., inequalities
107 √ √
36 36 6 Theoretical calculus

37 37(i) 1.5 Parabola, area of triangle 29

38 37(ii) 1.5 Circumradius of triangle
√ √
39 37(iii) 1.5 Centroid of triangle √ √ √
40 37(iv) 1.5 Circumcentre of triangle
S. Q. M Area(s) T C S I U F R
No. No.

41 38(i) 1.5 definite integral

42 38(ii) 1.5 area between parabolas
√ √
43 38(iii) 1.5 angle between curves √ √
44 38(iv) 1.5 differential equation
√ √
45 39(i) 1.2 coordinates, inequalities
√ √
46 39(ii) 1.2 vectors √
47 39(iii) 1.2 definite integrals

48 39(C) 1.2 definite integrals √ √ √
49 39(iv) 1.2 trigonometric optimisation
√ √ √
50 40(i) 1.5 trigonometric identity,
infinite sum √ √
51 40(ii) 1.5 solution of triangle
√ √
52 40(iii) 1.5 distance from a line √ √
53 40(iv) 1.5 distance from a plane 12,

Total counts 33 37 8 17 9 8 11

Setting a good JEE question paper is a tough examination where the

paper-setters are the candidates and the answer-book they write is the ques-
tion paper which they set for over three lakhs JEE aspirants, vying for every
single mark. Because of the keen competition, even a difference of a few
marks can mean the difference between being an electrical engineer and be-
ing a chemical engineer. And, if this happens because of flaws in the question
paper, that is most regrettable.
Fortunately, the 2006 JEE Mathematics question paper does not have any
serious mathematical mistakes. There are some obscure problems, such as
those in Comprehension II (about estimation of integrals). Sometimes, as in
Q. 39(i), the wording is confusing. Also, in Q. 39(iv), it is not given explicitly
that A, B, C are the angles of a triangle. But generally, anybody would take
this to be the case and so no harm is done.
But, even though there are no mathematical mistakes in the question
paper, there are many other undesirable features as the tables above show.
If the metaphor of the paper-setters being candidates at the paper-setting

examination is to be stretched further, then the tables above are evaluations
of their performance. Every tick mark in any of the columns marked under S
(sneaky answer), I (impropriety), U (unscrupulous) and F (familiar problem)
is like a red mark on the evaluated answer-book. For a good problem, all the
last five columns should contain blanks. In fact, just as a cartoon without
a verbal caption is considered as the best type of a cartoon, ideally a JEE
question should have only a blank in the column under C (computation)
too. That is, the computation involved should be so marginal that there is
virtually no possibility of going wrong because of computational mistakes.
Q. No.s 1, 16, 18, 22, 23, 30 and 36 are indeed of this type. Q. 28 would
also qualify had the line L in it been introduced properly. The paper-setters
deserve to be commended for setting these beautiful questions. Together
they carry 39 marks out of 184.
It is obviously too much to expect that all problems be of this type. Many
times computations are inevitable to give effect to the key idea. And as long
as the computations needed can be reasonably completed in the time allowed
(which is 40 seconds per mark as calculated earlier) there is no reason to
shun a problem simply because it involves some computation. Moreover, the
ability to carry out certain computations is also a prerequisite and has to be
tested somewhere.
In the present paper there are a few problems where the computations
can be reasonably completed in the time allowed, with some time left for
checking their accuracy. They include Q. No.s 5, 7, 10, 11, 14, 15, 27, 28, 29,
31, 33, 34 and 35. Together they account for 60 marks.
The rest of the questions, which add up to nearly half the credit, are
marred either because there is something confusing in the framing of the
question (thereby forcing the candidates to spend extra time just to see
what the question means) or because the computations involved are either
tricky or straightforward but far too lengthy. Glaring examples of the latter
type are Q. No.s 6, 8, 17, 19, 20, 21, 32, and all parts of Q. 37 to 40. It is
ridiculous to expect that the area of the region between two parabolas (Q.
38(ii)) can be evaluated in less than a minute.
The trouble arises because the persons who participate in setting a ques-
tion are usually poor judges of how long it takes to solve it. They have
already done the thinking part and expect the others to get the key idea in a
jiffy. They also often make the mistake of thinking that since no work needs
to be shown, very little time is needed for the computations. Also, since they
already know the correct answers to the computations, they probably think

that the very first attempt is free of any numerical mistakes and hence no
time is needed to check its accuracy. For most mortals this is far from the
To get a realistic idea of the time needed to answer a question completely,
the paper-setters would do well to try candidate simulation. This means
that one or two members of the paper-setting team should stay away com-
pletely while the remaining ones draft the question paper. Thereafter, these
two members should attempt the problems as if they are candidates, with
absolutely no help from those who have set the questions. Besides giving a
realistic idea of the time needed, this will also serve to expose ambiguities
and/or mistakes in the questions.
In terms of the topics covered, there are some avoidable duplications of
work, listed in the last column of the tables above. The objective type format
rules out the proofs of many identities and inequalities. Not surprisingly,
binomial identities (or even the binomial theorem for that matter) find no
representation in the entire paper. Trigonometric identities about a triangle
feature in Q. 3 and in Q. 15. But at both the places, those who can recollect
a relatively less known identity get an unfair advantage over those who have
to come up with it. In a conventional examination, the questions could have
given these identities and asked their proofs. In that case, the gap between
these two types of candidates would have narrowed down considerably.
Also totally absent are identities about vectors. As remarked in the com-
ments on Q. 20, a few of them could have been catered to, instead of asking
the candidates to do qualitatively same computations in all questions about
vectors. Among the conics, the ellipse and the hyperbola figure only once,
viz. in Q. 14. The parabola, on the other hand, comes up in Q. No.s 10, 13,
28, 29, 37 and again in 38(ii). This unevenness could have been corrected.
In the case of probability too, all the three problems asked (viz. Q. 21, 22
and 23) had the same underlying idea, viz. conditional probablility. Instead,
a question involving either a complementary or a binomial probability would
have been welcome.
All the four questions about matrices (viz. Q. 18, 30, 31, 32) are good. In
fact, two of them are among the best questions in the whole paper. Together
these four questions carry 20 marks, which is a disproportionately heavy
credit for a single topic. Two of these four questions (e.g. Q. 31 and 32)
could have been dropped to make some more room for number theory (which
is barely touched upon in Q. 7) or binomial identities (which are not catered
at all).

The overall picture is that speed will dominate all other qualities in the
selection. That also increases the importance of memory, previous prepara-
tion and, most importantly, the strategy. Although Q. 13 to 36 carry more
marks than Q. 1 to 12, there is not much difference in the level of difficulty
(except for those that are thought oriented). So those who attempted the
questions in the order they are asked stand to lose. The best strategy in a
paper like this would be to attempt Q. 13 to 36 first and then to give the
remaining time to Q. 1 to 12. If any time is left, try Q. 37 where the four
parts have a common theme. In each of the remaining three questions, (viz.
Q. 38 to 40), you have to answer four (or five) totally unrelated parts and
even if you make a single mistake, you get nothing. It is a costly gamble to
go for them.
The biggest disappointment comes from the comprehensions. This is elab-
orated at the end of the comments on Q. 32. Also, the author’s concept of a
comprehension is illustrated with a list of five hypothetical questions about
the second comprehension (Q. 24 to 26).

For JEE 2006, candidates were not allowed to take with them the question
papers. This is in sharp contrast with a comparable examination like the
AIEEE (meant for selection to the National Institutes of Technology) where
the candidates are allowed to take away the test booklet (which contains the
question paper). Nor were the JEE question papers displayed on the websites
of the IITs, as is done by some state boards that conduct the CET (Common
Entrance Test) for admission to the engineering colleges in their states. It is
shocking that examinations of similar types and comparable purposes should
differ so drastically in terms of their degree of transparency.
As a result, memory based versions of the questions had to be relied upon
while preparing the present commentary. Given the large number of questions
and the multiple answers to each, the text of the question paper became so
voluminous that it was humanly impossible, even with a concerted effort, for
any individual or any team to reproduce the full text verbatim. As a result,
many obscurities remained (e.g. Q. 10, 31, 32 and 38(iv)) as discussed in the
comments above.
It took more than two months and an intervention of the Right to Infor-
mation Act to get a photocopy of the JEE 2006 Mathematics question paper.
Many of the obscurities can now be resolved and the comments pertaining
to the respective questions have to be revised accordingly. But we have pre-
ferred to keep them as they are, as a grim testimony of the ill effects of a
needlessly secretive policy. Instead, we now give the official versions of the
questions and the consequent revisions in the comments. We omit the texts
where they substantially coincide with the memorised versions given above.
It may be noted that within each section, the questions are often permuted
among themselves in different versions of the same question paper. We stick
to the numbering in which the questions have appeared above. We shall only
change the text where it is substantially different and then comment on the
revised question.

Q. 1 No significant change.

Q. 2 No significant change.

Q. 3 No change except that instead of ‘incircle’ the words ‘circle inscribed’

are used. Also there is no mention of units, either for the length of the
radius or for the area.

Q. 4 No significant change. Hence the improprieties discussed earlier remain.

Q. 5 For 0 < θ < 2π, if 2 sin2 θ − 5 sin θ + 2 > 0, then θ lies in

(A) ( 5π
, 9π
) (B) ( π8 , 41π
(C) (0, 8 ) ∪ ( π6 , 41π
) (D) (0, 6 ) ∪ ( 5π
, 2π)

Comment: This wording is better because the word ‘interval(s)’ could

cause confusion when what is really meant is the union of intervals.
The incorrect answers given here are different than the earlier ones.
But that makes little difference because in this problem you cannot
get the correct answer by eliminating the wrong answers. The correct
answer remains the same except for the letter allotted to it.

Q. 6 Let w = α + iβ, β 6= 0 be a complex number. Then the set of complex

numbers z, (z 6= 1), such that
w − wz
is a real number
(A) {z : z 6= 1} (B) {z : z 6= 1 and |z| = 1}
(C) {z : |z| =
6 1} (D) {z : z = z}

Comment: This formulation is more direct. As in the last question,

some of the wrong alternatives are different. In a way that makes it
easier to do the problem in the sneaky way, i.e. by eliminating the
wrong answers. For example, the point z = 0 satisfies (A), (C) and (D)
all but not the given condition. So the correct answer must be (B). In
the earlier version, the correct answer is the same, except that its letter
was (D).

Q. 7 Let r, s, t be three distinct prime numbers. If p and q are two positive

integers whose least common multiple is r 2 s4 t2 , then the number of
pairs (p, q) is equal to

(A) 252 (B) 242
(C) 225 (D) 224

Comment: There are two changes in the text of the question: (i) the
three primes r, s, t are given to be distinct and (ii) the pairs (p, q) are
not given to be ordered pairs. In the earlier solution, we had assumed
(i) tacitly anyway. But the second change calls for some comment.
(p, q) is a standard notation for the ordered pair whose first entry is p
and the second entry is q. So, even though the word ‘ordered’ is not
used explicitly, one may take the question implicitly to mean that it
asks only for ordered pairs, In that case, the earlier solution will apply.
The standard notation for the unordered pair consisting of p and q is
{p, q}. So, if by (p, q) we mean the unordered pair, then the solution
will have to be modified accordingly. For example, in finding the pairs
of a and u in the solution, (2, 0) and (0, 2) represent the same pair.
Similarly, (2, 1) will equal (1, 2). Hence, instead of 5, we have only
three pairs, viz. (2, 0), (2, 1) and (2, 2). Similarly the number of pairs
(b, v) is not 9 but 5 while the number of pairs (c, w) would come down
from 5 to 3. Hence the answer to the problem would be 3 × 5 × 3,
i.e. 45. As this is not one of the given alternatives, we have to assume
that (p, q) stands for an ordered pair. But this way of arriving at the
intended interpretation is time-consuming to say the least. It would
have been far better to clearly specify in the text of the question that
(p, q) stands for an ordered pair.
Q. 8 No significant change. But the correct answer is listed as (B) instead of
(A). So, those who attempt the solution simply by differentiating the
given alternatives one by one, have to wait a little longer!
Q. 9 No substantial change.
Q. 10 A parabola has its axis along the line y = x with vertex and focus in
the √
first quadrant.
√ If the distances of vertex and focus from the origin
are 2 and 2 2 respectively, then the equation of the parabola is
(A) (x − y)2 = 8(x + y − 2) (B) (x + y)2 = 2(x − y + 2)
(C) (x − y)2 = 4(x + y − 2) (D) (x + y)2 = 2(x + y − 2)

Comment: This formulation is unambiguous and simpler too. It co-

incides with the second possible interpretation given earlier. So, the

equation of the parabola is (x − y)2 = 8(x + y − 2), i.e. (A). It is
strange that such a simply worded question was memorised with an
unnecessarily clumsy wording.

Q. 11 No significant change except that the function f (x) is defined by

 x, 0≤x≤1
f (x) = 2 − ex−1 , 1 < x ≤ 2
x − e, 2<x≤3

Z x
As before, g(x) is defined by g(x) = f (t)dt, 0 ≤ x ≤ 3. The four
alternatives given for the answer are same as before. (And, luckily,
their order is also the same.)

Comment: Because of the change in the definition of f (x), instead of

(1) in the earlier solution, we now have


 0≤x≤1
′ x−1
g (x) =  2 − e , 1 < x ≤ 2

x − e, 2<x≤3

Similarly, instead of (2) we now have


 0≤x<1
′′ x−1
g (x) = −e , 1 < x < 2
1, 2<x≤3

But these change do not affect the answer because the increasing/decreasing
behaviour of g(x) remains unaffected throughout the interval [0, 3]. So,
the correct answer is still (A). The last remark made in the earlier so-
lution also remains intact. That is, if in the question as it is given now,
the function f (x) were defined by f (x) = ex−1 − 2 instead of 2 − ex−1
on the interval (1, 2], then the problem would have been far more in-
teresting, because in that case f (x) would have changed its sign at all
the four points 1, 1 + ln 2, 2 and e and both (A) and (B) would have
been correct.

Q. 12 Let ~a = î + 2ĵ + k̂, ~b = î − ĵ + k̂ and ~c = î + ĵ − k̂ be three vectors. Then
a vector
√ in the plane of ~a and ~b whose projection on ~c is of magnitude
1/ 3 is
(A) 2î − 3ĵ − 2k̂ (B) 4î − ĵ + 4k̂
(C) 4î − 7ĵ − 4k̂ (D) 3î − 5ĵ − 3k̂

Comment: The alternatives given for the answer are different (numer-
ically) from those given earlier. But the data is the same, except for the
difference that we are now only given the magnitude of the projection
on the vector ~c and not the projection
√ itself. As a√result, the actual
projection can also equal −1/ 3 instead of only 1/ 3. This does not
affect most of the working. In particular, if we take a desired vector as
~v = v1 î + v2 ĵ + v3 k̂ then we get v1 = v3 exactly as before. It is ironic
that among the four given alternatives only (B) satisfies this condition.
So, without doing anything further, a candidate can mark it as the cor-
rect answer! Note that here we have only used that ~v lies in the plane
of ~a and ~b. The part involving its projection on ~c does not come into
the picture at all! To avoid this sneaky path, the paper-setters ought
to have included at least one more alternative where the coefficients of
î and k̂ are equal.
For an honest answer, we must not, of course, capitalise on such lapses
on the part of the paper-setters. Instead, let us see what the condition
about the projection of ~v on ~c means. As in the earlier solution, the
~v · ~c
projection of ~v on ~c is √ . But we can no longer set this equal to
√ 3
1/ 3. Instead, we first take its magnitude, i.e. its absolute value and
√ ~v · ~c 1
set it equal to 1/ 3. In other words, we get √ = ± √ , and hence
3 3
v1 + v2 − v3 = ±1 instead of the equation v1 + v2 − v3 = 1 obtained
earlier. As before, since we already know that v1 = v3 , this equation
simply means v2 = ±1. Among the given alternatives, (B) is the only
one where this condition is satisfied.

Q. 13 Tangents common to both the parabolas y = x2 and y = −x2 + 4x − 4

(A) y = 4(x − 1) (B) y = 0
(C) y = −4(x − 1) (D) y = −30x − 20

Comment: There is no substantial change, except that the in the

equation of the second parabola, we now first have to complete the
square. The purpose of adding such complications is far from clear.
Surely, they are not relevant to the main theme of the problem. They
only add to its drudgery. But then, possibly as a compensation, the
paper-setters have also been kind enough to tell us that there are more
than one tangents, instead of the earlier version which was fussily non-
committal in this respect.

Q. 14 No significant change, √ except that in the

√ last alternative, the focus
of the parabola is (3 5, 0) instead of (5 3, 0). But that makes little
difference because anyway it is a wrong alternative and in a problem
like this you can’t get the answer merely by eliminating the wrong
alternatives sneakily.

Q. 15 No significant change, except that in the statement of the problem, the

line through D perpendicular to AD is given to meet AC in E and AB
produced in F . In other words, the word ‘produced’ is added. This is
an old practice in geometry where a fussy distinction was made between
a point on the side AB and a point on the side AB produced, the latter
indicating that the point lies on the line AB but not on the segment
AB. Sometimes the phrase ‘produced if necessary’ was used to allow
both the possibilities. Nowadays, people are not so fussy about these
trifles. In fact, in the present problem, the given possibility occurs
when b > c, i.e. when the side AC is bigger than the side AB. If it
were the other way, then the perpendicular to AD through D will meet
the side AB and the side AC produced. But all the four alternative
answers are independent of which of these two sides is bigger. So there
was really no need to take sides here. A perfectly safe wording which
encompasses both the possibilities would have been to simply say that
the line through D perpendicular to AD meets the lines AC and AB
in points E, F respectively.

Q. 16 No significant change.

Q. 17 Let f (x) be a polynomial of degree 3 having a local maximum at x =

−1. If f (−1) = 2, f (3) = 18, and f ′ (x) has local minimum at x = 0,

(A) the distance between (−1, 2) and (a, f (a)), which are the points
of√local maximum and local minimum on the curve y = f (x) is
2 5

(B) f (x) is a decreasing function for 1 ≤ x ≤ 2 5

(C) f ′ (x) has a local maximum at x = 2 5
(D) f (x) has a local minimum at x = 1

Comment: Qualitatively, the nature of the problem remains the same,

viz. to determine a cubic polynomial from four pieces of data (because
there are four unknowns associated with a general polynomial of degree
3) and then to answer questions about the known cubic.
But the numerical data is somewhat different than the earlier version.
As a result, some of the calculations will change, although not in spirit.
Exactly as before, because of the first and the last piece of the data, we
take the cubic as f (x) = ax3 − 3ax + d. But after this the calculations
change numerically because the other two given conditions are slightly
different. So, instead of the equations in (1) of the earlier solution, we
now have
2a + d = 2 and 18a + d = 18
solving which we get a = 1 and d = 0. These values are, in fact, simpler
than the ones we obtained earlier. So, the polynomial f (x) is simply
x3 −3x. This gives f ′ (x) = 3x2 −3x and f ′′ (x) = 6x−3, from which we
further get that f has a local maximum at x = −1 (which is given to us
anyway) and a local minimum at x = 1 (which is new to us). So, we get
a = 1 and (a, f √(a)) = (1, f (1))
√ = (1, −2). The distance of this point
from (−1, 2) is 4 + 16 = 2 5. So the statement (A) is true. From
f ′ (x) = 3x(x − 1) we√see that f (x) is increasing on [1, ∞) and hence,
in particular on [1, 2 5].√ So (B) is false. As for (C), f ′′ (x) = 6x − 3
does not vanish at x = 2 5.√So f ′ (x) has neither a local minimum nor
a local maximum at x = 2 5. Hence (C) is also false. The truth of
(D) is already known.

Q. 18 No significant change.

Q. 19 Let C be a curve such that the tangent at any point P on it meets the
x-axis and the y-axis at A and B respectively. If BP : P A = 3 : 1 and
the curve passes through the point (1, 1), then

(A) The curve passes through (2, 1/8)

(B) Equation of normal to the curve at (1, 1) is 3y − x = 2
(C) The differential equation for the curve is 3y ′ + xy = 0
(D) The differential equation for the curve is xy ′ + 3y = 0

Comment: The question is basically the same. But the alternatives

are slightly different. The wording of the data is a little faulty. The
first sentence does not specify any property of the curve. It would do
so only if the points A and B were introduced earlier. Instead, what
you are saying is that you are considering a typical point, say P , on the
curve C and introducing A and B as the points where the tangent at P
meets the x- and the y-axes respectively. You can do this to any curve.
So, it is wrong to use phrases like ‘such that’ or ‘with the property that’
in a sentence like this. This is as silly as saying, “Let A be a man such
that his father is F .” A better wording is to say, “Let A be a man and
let F be his father.”
In the given question, it is only the second sentence which specifies
the vital properties which characterise the curve. A better wording
would have been
“Let C be a curve with the property that whenever the tangent
at any point P on it meets the x- and the y- axes at points A and B
respectively, BP : P A = 3 : 1. Suppose also that C passes through the
point (1, 1).”
In this formulation, the first sentence implies that the curve C
satisfies a certain geometric property. This property makes the curve
belong to a certain class of curves (specifically, to a one parameter
family of curves). The second sentence then tells us which member of
this class is the given curve C. If the first sentence in this formulation
appears too complicated, here is a simpler one where the contents of
the first sentence are split into two.
“Let P be a variable point on a curve C and let the tangent to
C at P meet the x-axis at A and the y-axis at B. Suppose, for every

position of the point P on the curve, BP : P A = 3 : 1. If further, C
passes through the point (1, 1), then”
While the given formulation is not likely to cause any confusion,
one hopes that a certain linguistic discipline is followed in framing the
text of a question.
Mathematically, the data is exactly the same as before and so the
working also remains the same. We already derived xy ′ + 3y = 0 as
the differential equation satisfied by the curve C. So (D) is a correct
statement and (C) is false. We also solved this d.e. earlier and showed
that (A) is true. But the alternative (B) here is different than the
earlier. The slope of the normal at (1, 1) is 31 as calculated earlier.
Hence the equation of the normal is y = 31 x + 23 . Thus we see that (B)
is true now.

Q. 20 No significant change. (Hats off to persons who can correctly remember

so much numerical data when the particular numbers occurring in it
are essentially arbitrary and follow no logical pattern.)

Questions from Q. 21 to Q. 32 have been grouped into four groups. Each
bunch of three consecutive questions is preceded by a ‘paragraph’ to be read
first. (Earlier we called it the ‘preamble’.) The paragraph is supposed to
give the common setting for the three questions appearing under it.

The paragraph for Q. 21 to Q. 23 reads :

“ There are n urns numbered 1, 2, . . . , n each containing (n + 1) balls. Urn

i contains i white balls and (n + 1 − i) red balls, i = 1, 2, . . . , n. An urn is
selected and a ball is drawn at random from it. Let Ui denote the event that
urn numbered i is selected and let W denote the event that a white ball is
drawn from the selected urn. Further suppose that E denotes the event that
an even numbered urn is selected.”

Q. 21 No significant change.

Q. 22 No significant change.

Q. 23 No significant change.

Comment: The ‘paragraph’ above is worded in a language which is easy to

understand. But there is little ‘comprehension’ in it because it introduces
no new concept. The three problems are like any other problems and the
‘paragraph’ is merely a statement of the hypothesis.

The paragraph for Questions from 24 to 26 is a little more detailed. Also,

there is indeed something conceptual in it, viz. approximation of a definite
integral. It reads as follows:

“Let y = f (x) be a twice differentiable, non-negative function defined on

Z b
[a, b]. The area f (x)dx, b > a bounded by y = f (x), the x-axis and the
ordinates at x = a and x = b can be approximated as
b (b − a)
f (x)dx ≈ {f (b) − f (a)}.
a 2
Z b Z c Z b
Since f (x)dx = f (x)dx + f (x)dx, c ∈ (a, b), a better approxi-
aZ a c
mation to f (x)dx can be written as
Z b (c − a) (c − b)
f (x)dx ≈ {f (a) + f (c)} + {f (c) + f (b)} ≡ F (c).
a 2 2
If c = , then this gives :
Z b
f (x)dx ≈ {f (a) + 2f (c) + f (b)}. (1)
a 4
Z π/2
Q. 24 The approximate value of sin x dx using rule (1) given above is
π √ π √
(A) √ (1 + 2) (B) √ (1 + 2)
8 2 4 2
π √ π √
(C) (1 + 2) (D) (1 + 2)
8 4
Comment: Now the question is unambiguous because it clearly tells
you which formula to apply.

Q. 25 No significant change.
Q. 26 No significant change. But now the ambiguity is gone because the
Z b
symbol f (x)dx means only one thing, viz. the integral of f (x) over
[a, b].

The paragraph for Q. 27 to 29 reads :

“ The length of a side of a square ABCD is 2. A circle C1 inside the square

touches all the four sides of the square and another circle C2 passes through
all the four vertices of the square. Let P and Q be any two points on the
circles C1 and C2 respectively. A point S moves in the coordinate plane such
that it is always at an equal distance from a fixed line L and a fixed point
R. A circle C in the coordinate plane always touches the circle C1 externally
and the line L.
Q. 27 No significant change.
Q. 28 Let L be the line passing through any two adjacent vertices of the
square. If circles C1 and C are on the same side of the line L, then the
locus of the centre of C is
(A) an ellipse (B) a hyperbola
(C) a parabola (D) a pair of straight lines

Comment: The same objections given earlier apply here. It is unfair

to introduce the line L in the paragraph without any restriction and
then put a restriction on it later. However, the particular restriction on
L given here simplifies the problem slightly and makes the co-ordinate
geometry solution more lucrative. (The pure geometry solution under-
goes little change.)
Without loss of generality, we take L to pass through the vertices A
and B. So, if we set up the coordinate system as in the solution to Q.
27, the equation of L comes out to be y = 1. Let (h, k) be the centre
of the variable circle C and r its radius. As C and C1 lie on the same
side of L, we have k ≤ 1 and so the perpendicular distance of (h, k)
from L is 1 − k. Since C is given to touch L, we have
r =1−k (1)

But C also externally touches C1 whose centre is (0, 0) and radius 1.
Therefore we also have

1 + r = h2 + k 2 (2)

The desired locus is obtained by eliminating r from (1) and (2). It

comes out as

2 − k = h2 + k 2 (3)

which simplifies to

h2 = 4 − 4k (4)

which is a parabola.

Q. 29 Let the point R be at A and L be the line passing through the vertices
B and D, adjacent to A. If the locus of the point S intersects AC at
T1 and the line passing through A parallel to L at T2 and T3 , then the
area of the triangle T1 T2 T3 is
(A) 4 (B) 1
(C) 3 (D) 2

Comment: There is no mention of units for the area. The definition

of the line L given here is different from that in the last question. Also
the manner in which it is given is unnecessarily clumsy. It would have
been easier to say that let L be the line along the diagonal BD. Also
it is not clear what is gained by introducing R as a ‘fixed point’ in the
paragraph, and by identifying it with A now. Neither of the other two
questions make any reference to the point R. So it would have been
far better to simply drop any mention to it in the paragraph and word
the present question suitably.
The contents of the question remain the same and so does the

The paragraph for Questions from 30 to 32 reads:

 
2 0 1
“ Let A =  0 1 −1 . Suppose U1 , U2 , U3 are three column vectors
 

1 1 0
such that      
1 2 3
AU1 =  0  , AU2 =  3  and AU3 =  2
     

0 0 1
U is a 3 × 3 matrix whose first, second and third columns are U1 , U2 and
U3 respectively. ”

Q. 30 The value of the determinant of U is

(A) 2 (B) 3
(C) 6 (D) 12

Comment: The phrase ‘column matrices’ used earlier has been re-
placed by the more customary ‘column vectors’. Conceptually the
problem is the same. But the entries of the matrix A as well as those
of AU3 are different from the earlier ones. As a result, most of the
computations have to be redone. For the present question, we follow
the short cut of finding the determinant of U without first finding U
itself. We are given that
 
1 2 3
AU =  0 3 2  (1)
 

0 0 1

from which we readily compute |AU| = 3 quickly because this is an

upper triangular matrix. With A we are not so lucky. Still, by straight
expansion we see that its determinant is 1. Substituting these values
in the equation

|AU| = |A||U| (2)

we get |U| = 3. This is the same answer as before. But that is a

coincidence because the data is now numerically different.

Q. 31 The sum of the elements of U −1 is

1 1
(A) (B)
12 6
1 1
(C) (D)
3 4

Comment: Once again, we try to avoid having to compute U. As in

the earlier solution, if we let
 
1 2 3
B= 0 3 2  (3)
 

0 0 1
then (1) reads as AU = B. Hence U = A−1 B. Taking the inverses of
both the sides we get
U −1 = B −1 A (4)
and hence the desired sum, say S, of all entries of U −1 is given by
 
S = [1 1 1] B −1 A  1  (5)
 

which is exactly same as (17) in the earlier solution except that the
matrices A and B are different now. So we have do the computations
over again. But luck is still with us, because the matrix B remains
upper triangular and therefore its inverse can be computed a lot more
efficiently than that of a general matrix. We skip the computations
and only write down the result, viz.
 
1 −2/3 −5/3
B −1 =  0 1/3 −2/3  (6)
 

0 0 1
As before, instead of computing B −1 A (which would give us U −1 ), we
use associativity of matrix multiplication in (5) and calculate P and Q
where where
P = [1 1 1] B −1 (7)
 
and Q = A  1  (8)
 

both of which can be done by inspection from
 the entries of B and
A. Thus P = [1 − 1/3 − 4/3] and Q =  0 . Multiplying the two
 

we finally get S = 3 + 0 − (8/3) = 1/3. Hence the correct answer is
 
 2  is
Q. 32 The value of [3 2 0]U  

(A) 13 (B) 26
(C) 12 (D) 24

Comment: As observed in the comments on the earlier version of the

question, there would be an elegant way to do this problem if B −1 were
the transpose of the matrix A. For, in that case A−1 would equal B t ,
the transpose of B. Since
 U = A B, the desired matrix product would
equal [3 2 0]B t B 
 2 , which can be evaluated simply by computing

the row vector [3 2 0]B t and then taking the sum of the squares of its
But unfortunately, we do not have B −1 = At . So there is apparently
no elegant way to do the problem. The only simplification
  we can do
is to rewrite the expression asked as [3 2 0]A−1 B  2  and hence as
 

the product LM where L is a row vector and M is a column vector,
defined respectively by

L = [3 2 0]A−1 (9)
 
and M = B 2  (10)
 

There is no elegant way to compute A−1 either because the matrix A

is not of a particularly simple type such as a triangular matrix. So,
we compute A−1 by taking the transpose of the matrix of the cofactors

of its entries and dividing by the determinant |A|, which, fortunately
equals 1. The answer comes out to be

 
1 1 −1
A−1 =  −1 −1 2  (11)
 

−1 −2 2

A direct computation gives

 
1 1 −1
L = [3 2 0]  −1 −1 2  = [1 1 1] (12)
 

−1 −2 2
    
1 2 3 3 7
and M =  0 3 2   2  =  6  (13)
    

0 0 1 0 0

and hence, finally,

   
3 7
[3 2 0]U  2  = LM = [1 1 1]  6  = 7 + 6 + 0 = 13 (14)
   

0 0

So the correct answer is (A).

As there is no elegant way to answer this question, the work
becomes laborious and very prone to numerical errors. Also, it is highly
repetitious. When you multiply two matrices of order 3 or find the
inverse of a general 3 × 3 matrix, you have to do qualitatively same
computation 9 times. The purpose of testing this ability is questionable.
Comparatively, the computations in the last question were not such
a drudgery if you followed the elegant approach. Still, in terms of
proportionate time (viz. just about 3 minutes for a 5 points question),
they were too many. It is only Q. 30 which can truly be called a good

Q. 33 No significant change, except that a, b, c, d are given to be real numbers,
which is nowhere needed in the solution.

Q. 34 No significant change.

Q. 35 No significant change.

Q. 36 No significant change.

Q. 37 No significant change.

Q. 38 There is no change in the first three entries of the first column and the
entries (A), (B) and (D) of the second. But entry (C) is given as 2 log 6
instead of 6 ln 2. Item (iv) of the first column reads :
“ A continuous function f : [1, 6] −→ [0, ∞) is such that f ′ (x) =
and f (1) = 0. Then the maximum value of f cannot exceed
x + f (x)

Comment: The differential equation satisfied by the function f (x)

is the same as before (except for the change of notation). The initial
condition is also the same. So, the solution also remains the same, viz.

x + f (x) + 2 = 3ef (x)/2 (1)

The problem involves the maximum of the function f (x) over the in-
terval [0, 6]. It is impossible to solve (1) explicitly for f (x). Nor is
it needed. From the statement of the question we have to assume
that the denominator x + f (x) never vanishes in the interval [0, 6].
So, by continuity, it has to be either positive throughout or negative
throughout. The latter possibility is excluded because we are given that
1 + f (1) = 1 > 0. So we conclude that x + f (x) and hence also f ′ (x)
are positive throughout the interval [0, 6]. Therefore f (x) is strictly

increasing on [0, 6]. Hence the maximum of f (x) on [0, 6] occurs at
x = 6. This maximum, say M (= f (6)) is given by

M + 8 = 3eM/2 (2)

Again, from this equation we cannot determine M explicitly. But that

is not necessary either. All we are asked is to find which of the four given
alternatives in the second column of the question are upper bounds on
M. These alternatives (in ascending order) are 0, 1, 4/3 and 2 ln 6. (We
are not given an approximate value of ln 6. However, for the purpose
of the present comparison, it is enough that ln 6 > 1 which is true since
6 > e.)
The problem now boils down to deciding between which adjacent
pair M lies. We have no way of computing the number M even approx-
imately from (2). So, it will not be easy to compare M directly with
other numbers. But we know that M = f (6) and since f (x) is known
to be strictly increasing it is easier to compare M with the values of f
at other points. Let us see if there is any value of x in the interval [0, 6]
at which f (x) = 2. If we can show that such a value exists, then it will
follow that M ≥ 2. To look for such x we put f (x) = 2 in (1) and solve
the resulting equation for x to get x = 3e − 4. Since 2 < e < 3, we see
that this number lies between 2 and 5. So, it is in the interval [0, 6)
and therefore, as we just said, we know M > 2, which automatically
implies that M is bigger than all the three numbers 0, 1 and 4/3 in the
second column. Hence none of these three is a correct match for the
entry (iv) in the first column of the question.
It only remains to compare M with the number 2 log 6. So, let us
see if there is any value of x for which f (x) = 2 log 6. Once again we
put f (x) = 2 ln 6 in (1) and solve to get x = 16 − 2 log 6. Even though
we do not know an approximate value of log 6, from the knowledge that
2 < e we know that 6 < 8 < e3 and hence taking natural logarithms,
log 6 < 3. Therefore, 16−2 ln 6 > 16−2×3 = 10. It follows that there is
no value of x in [0, 6] for which f (x) = 2 log 6. From this we claim that
M < 2 log 6. For, otherwise we shall have 2 log 6 ≤ M = f (6). Since
we also have 2 log 6 > 0 = f (1), by the Intermediate Value Property,
there would be some point in the interval [1, 6] at which f would equal

exactly 2 ln 6. But we just showed that there is no such point even in
the larger interval [0, 6].
So, finally we have proved that the maximum value of f on [0, 6]
cannot exceed 2 log 6. Hence the correct match for (iv) in the fist
column is (C) in the second column. The reasoning needed to arrive at
this is non-trivial and that makes the problem an excellent one in an
examination where the ability to reason is tested. But in the present
case, chances are that most students will simply not bother about this
reasoning and will tick the answer 2 log 6 on the superficial ground that
since the solution of the differential equation involves the exponential
function, there is a good chance that a problem about it would have
natural logarithms in it. Such gamblers will actually be rewarded. Yet
another instance where a good problem is marred by the multiple choice
format of the examination.
Q. 39 There is no change in the entries in the second column (except for their
relative order). In the first column, (ii), (iii) and (iv) are the same.
But there is a slight change in the first two entry. It reads :
“ The set of values of a for which the lines

x + y = |a|, ax − y = 4,

intersect in the region x > 0, y > 0, is the interval (a0 , ∞). Then the
value of a0 is ”

Comment: This formulation is much better than the earlier one be-
cause it avoids the word ‘ray’. There is a slight change in the numerical
data, viz. the R.H.S. of the equation of the second line is 4 instead of
1. So the calculations have to be redone. But the method remains the
same. Solving the equations together, we get
|a| + 4
x = (1)
a|a| − 4
and y = (2)
as the coordinates of their point of intersection. As before the positivity
of x implies a + 1 > 0, i.e. a > −1. In view of this, the positivity of y
means that a|a| > 4. This rules out negative values of a and therefore

a|a| is simply a2 . So we have a2 > 4 or equivalently, a > 2 as a is
already known to be positive. So a0 = 2. Thus the correct match for
(i) is (A) and not (B) as in the earlier version.

Q. 40 Item (A) in the second column is 0 instead of 2 2. In the first column,
there is no substantial change in the first three entries. But the last
entry is a little different numerically. It reads :
(iv) Let P be the plane passing through the point (2, 1, −1) and per-
pendicular to the line of intersection of the planes√2x + y − z = 3 and
x + 2y + z = 2. Then the distance from the point ( 3, 2, 2) to the plane
P is ”

Comment: As the changes are only numerical, we spare the details.

To get the equation of the plane P we follow the second approach,
based on the cross product of the normals to the two given planes.
These normals are 2î + ĵ − k̂ and î + 2ĵ + k̂ respectively. Their cross
product is
î ĵ k̂

2 1 −1 = 3î − 3ĵ + 3k̂

1 2 1

So, a normal vector to P can be taken as î − ĵ + k̂. As the plane passes

through (1, 1, −1), its equation is

(x − 2) − (y − 1) + (z + 1) = 0

i.e. x − y + √z = 0. Hence its perpendicular distance from the point

√ 3−2+2
( 3, 2, 2) is √ = 1. So the correct match for (iv) in the first
column is (B). (This is also the correct match for (i). But the preamble
to the Section V makes it clear that this is permissible.)